Social profiling

Fact-checked by Grok 2 weeks ago

Social profiling refers to the systematic collection and analysis of data from individuals' online activities, particularly on social media, to infer attributes such as demographics, personality traits, interests, and behavioral patterns, often employing machine learning algorithms to construct detailed user profiles.^[1] This process aggregates publicly available or user-generated content, including posts, images, and network connections, to enable predictions about unobserved characteristics. Key applications span marketing, where profiles facilitate targeted advertising by segmenting consumers based on psychographic data; security, including detection of fake accounts and potential threats through anomaly identification; and recruitment, assessing candidates' social influence and suitability via AI-assisted screening.^[2]^[3]^[4] Empirical evaluations of these techniques reveal substantial predictive power, with machine learning models achieving accuracies of 78-87% in forecasting personality types from platform data and up to 96% in distinguishing genuine from fraudulent profiles.^[5]^[6] While controversies center on privacy erosion, re-identification risks, and allegations of perpetuating biases—such as in facial analysis tools showing disparities across demographics—rigorous studies underscore that profiling's efficacy stems from leveraging statistically valid correlations in social data, countering claims of predominant inaccuracy by demonstrating that group-based inferences often align closely with observed realities rather than illusory prejudices.^[7] Such findings highlight the causal utility of data-driven profiling in real-world decision-making, though ethical deployment requires balancing informational benefits against surveillance concerns.^[8]

Definition and Fundamentals

Core Definition and Processes

Social profiling is the computational process of extracting and inferring users' personal attributes—such as demographics, psychographics, interests, behaviors, and preferences—from data generated on social media platforms, including textual posts, images, interactions, and network connections.^[9]^[10] This approach leverages user-generated content and relational data to construct individualized or aggregated profiles, enabling predictions about future actions or categorizations beyond explicit self-disclosures. The core processes of social profiling unfold in sequential phases, beginning with data collection, which aggregates raw inputs from platforms like Twitter or Facebook via APIs, scraping, or user permissions, encompassing profiles, timelines, likes, shares, and follower graphs.^[9] Preprocessing follows, involving cleaning noisy or informal language, handling missing values, and normalizing heterogeneous data types such as text, metadata, and multimedia.^[10] Feature extraction then identifies salient signals, applying techniques like tokenization for text, edge detection for images, or centrality measures for networks to quantify patterns in sentiment, topics, or connectivity.^[9] Inference modeling constitutes the analytical core, employing machine learning algorithms to map features to attributes; for instance, support vector machines classify gender from linguistic cues, while deep convolutional neural networks achieve up to 95% precision in age group prediction from visual content, and clustering methods group users by shared behavioral correlations.^[10]^[9] Individual profiling targets singular traits, such as inferring Myers-Briggs personality types from post patterns with varying accuracy across dimensions (e.g., lower for thinking-feeling axes due to data subtlety), whereas group profiling aggregates signals to characterize collectives, using motif analysis on interaction graphs to detect community trends.^[10] Performance evaluation assesses model outputs against ground-truth labels, often revealing limitations from data sparsity or cultural variances in expression. These steps enable scalable profiling but hinge on the causal linkages between observable social signals and latent traits, grounded in empirical correlations rather than deterministic causation.^[9]

Distinctions from Data Profiling and Psychological Profiling

Social profiling specifically involves inferring an individual's social attributes—such as interests, affiliations, network connections, and behavioral tendencies—from publicly available data generated through online social interactions, posts, and relationships on platforms like social media.^[11]^[12] This process emphasizes relational and communicative patterns, often for applications like personalized recommendations or targeted marketing, distinguishing it from broader data analysis techniques.^[13] In contrast, data profiling refers to the systematic examination of datasets to assess structure, quality, content, and patterns, such as identifying data types, duplicates, or inconsistencies within databases or records.^[14] This method prioritizes technical metadata and aggregate statistics for data governance and integration purposes, without a primary focus on deriving individualized social or behavioral inferences from interactive online sources. While social profiling may employ data profiling tools as a preparatory step to clean social media datasets, the end goal diverges: data profiling supports data usability, whereas social profiling constructs user-centric social models.^[14]^[13] Psychological profiling, meanwhile, entails deducing personality traits, motivations, cognitive styles, or behavioral predispositions, typically through forensic analysis of crime scenes, clinical interviews, psychological tests, or observed actions in controlled settings.^[15]^[16] Employed in criminal investigations or human resources, it relies on inductive reasoning from behavioral evidence to categorize offender types or individual psychologies, rather than aggregating digital social traces like likes, shares, or follower graphs.^[15] Unlike social profiling's reliance on passive, voluminous online social data for probabilistic social inferences, psychological profiling demands deeper interpretive analysis of intentional or symptomatic behaviors, often validated against established psychological frameworks.^[16]^[13]

Historical Development

Origins in Early Digital Tracking

Early digital tracking originated with the basic logging capabilities of web servers in the early 1990s, which recorded IP addresses, timestamps, user agents, and HTTP referrers to analyze aggregate site traffic and visitor patterns. These server logs enabled rudimentary insights into user navigation but lacked mechanisms for persistent individual identification, limiting profiling to statistical summaries rather than personalized behavioral models.^[17] The release of the first commercial web analytics tool, Webtrends, in 1993 further formalized log-file analysis, processing data to filter out bots and identify traffic sources, though it still relied on anonymized aggregates without cross-session linkage.^[17] A transformative shift occurred in 1994 with the invention of HTTP cookies by Netscape engineer Lou Montulli, initially designed to maintain stateful sessions for e-commerce features like shopping carts. Cookies stored small data snippets on users' browsers, allowing websites to recognize returning visitors, track session durations, and personalize content based on prior interactions—foundational elements for constructing individual user profiles from browsing histories.^[18] This persistence enabled inference of interests and behaviors, evolving from simple session management to behavioral tracking across visits.^[19] By the mid-1990s, third-party tracking emerged, exemplified by the founding of DoubleClick in 1996, which established the first internet advertising network brokering targeted ads via cookie-based monitoring of user actions across multiple sites.^[20] Web beacons, or invisible tracking pixels, also appeared in the late 1990s, embedding one-pixel images that triggered data transmission to remote servers upon page loads or email opens, facilitating cross-domain surveillance without direct user navigation.^[21] These technologies collectively pioneered the aggregation of digital footprints into inferential profiles, setting the stage for more sophisticated social data integration by enabling scalable observation of online interactions akin to social behaviors.^[18] The expansion of social profiling accelerated with the emergence of dedicated social media platforms in the late 1990s and early 2000s, which provided unprecedented volumes of user-generated data on interpersonal connections, behaviors, and preferences. Six Degrees, launched in 1997, represented an early precursor by enabling users to create profiles and form lists of connections, laying groundwork for network-based inference, though its short lifespan limited broader impact.^[22] Subsequent platforms like Friendster in 2002 and MySpace in 2003 introduced customizable profiles with interests, photos, and friend lists, allowing initial forms of social attribute deduction such as influence within peer groups or cultural affiliations derived from shared music and blog content.^[23] Facebook's 2004 debut catalyzed a pivotal shift, as its emphasis on real-name verification and expansive "social graph"—encompassing friend relationships, wall posts, and event participations—facilitated algorithmic mapping of users' social capital, ideological leanings, and lifestyle patterns at population scale. By 2006, Twitter's introduction of microblogging and hashtag systems added temporal and topical data layers, enabling real-time profiling of public sentiment and affiliation networks through analysis of follows, retweets, and conversational threads. These platforms' APIs, which granted programmatic access to such data until mid-2010s restrictions, empowered developers and researchers to aggregate and infer traits like personality types or political orientations from interaction patterns, with studies documenting over 1,000 Scopus-indexed publications on social media user profiling between 2012 and 2022 alone.^[9]^[12] This era's growth intertwined with rising platform user bases—Facebook surpassing 1 billion monthly active users by 2012—and the integration of behavioral tracking cookies, which linked social data to off-platform activities for holistic profiles. Commercial entities exploited these resources for targeted interventions, as evidenced by early advertising models where inferred demographics from likes and shares drove ad personalization, yielding measurable lifts in engagement rates of up to 20-30% in controlled experiments. However, expansions also amplified risks, with documented cases of data brokers compiling shadow profiles from public posts, often without explicit consent, underscoring the platforms' role in democratizing yet commodifying personal social metadata.^[9]^[24]

Key Milestones and Recent Advancements

The launch of SixDegrees.com in 1997 represented an early milestone in social profiling, as the platform enabled users to create personal profiles and form connections based on shared attributes, facilitating basic aggregation of social data for network analysis.^[25] This was followed by the advent of more sophisticated platforms; Facebook's 2004 debut introduced detailed explicit user profiles encompassing demographics, relationships, and interests, which served as foundational datasets for inferring behavioral patterns.^[26] By 2007, Facebook implemented targeted advertising leveraging these profiles, allowing advertisers to segment audiences by self-reported data such as location, education, and hobbies, marking the commercialization of social profiling for marketing precision.^[27] A pivotal advancement occurred in the mid-2010s with the integration of implicit behavioral signals. The 2010 rollout of Facebook's Open Graph protocol expanded data collection by tracking interactions across external sites via "Like" buttons and shares, enabling inference of latent preferences beyond explicit inputs.^[9] This era saw the rise of psychographic profiling, exemplified by Cambridge Analytica's 2014-2015 harvesting of data from approximately 87 million Facebook users through a personality quiz app developed by Aleksandr Kogan, which extrapolated psychological traits using the OCEAN model from friends' networks without consent.^[28] The firm's application of these profiles for micro-targeted political ads in the 2016 U.S. presidential election highlighted both the potency and ethical perils of inferred social profiling, though subsequent analyses questioned the causal efficacy of psychographic targeting due to limited predictive validity in real-world behavioral influence.^[29] The 2018 scandal's exposure prompted regulatory responses, including the EU's GDPR enforcement in May 2018, which imposed consent requirements and data minimization for profiling activities.^[30] Recent advancements from 2020 onward have centered on AI-driven enhancements to profiling accuracy and scale. Machine learning models, particularly graph neural networks, have improved inference from social graphs, predicting user attributes like political leanings or purchase intent with higher precision by analyzing interaction patterns across platforms.^[31] For instance, post-2021 iOS privacy updates limiting cross-app tracking, platforms adapted via federated learning techniques, processing data on-device to derive profiles without central aggregation, as evidenced in Meta's 2023 updates to recommendation algorithms.^[32] By 2024, large language models enabled multimodal profiling, integrating text, images, and video for nuanced trait extraction, such as sentiment-based personality assessment, boosting engagement prediction accuracy by up to 20-30% in controlled studies.^[9] The EU AI Act, effective August 2024, classifies certain social profiling uses as high-risk, mandating transparency and human oversight, while spurring innovations in privacy-preserving methods like differential privacy to balance utility and compliance.^[24] These developments underscore a shift toward causal inference models that prioritize empirical validation over correlative assumptions, amid ongoing debates over bias amplification in AI-derived profiles from unrepresentative training data.^[31]

Technical Components

Social data utilized in profiling originates predominantly from online social networks, where users voluntarily share personal information, interactions, and metadata. These platforms aggregate vast quantities of user-generated content, including textual posts, images, videos, and multimedia shares, which reveal linguistic styles, expressed opinions, affiliations, and temporal behaviors such as posting frequency and timing. For instance, platforms like Facebook, Instagram, and X (formerly Twitter) enable the extraction of such data through public APIs or scraping methods, yielding datasets with millions of entries for analysis.^[9]^[33] User profile attributes form a core static source, encompassing self-reported or inferred details like demographics (e.g., age, gender, location), interests, education, and employment history. These elements, often incomplete or outdated due to user privacy settings, provide baseline identifiers for matching across datasets. Interactional data, including likes, comments, shares, follows, and mentions, captures dynamic behavioral patterns, such as reciprocity in networks or engagement with specific topics, enabling inference of social capital or influence. Network structures—edges representing friendships, follows, or group memberships—offer relational insights, quantifying centrality measures like degree or betweenness to map community roles.^[12]^[31]^[1] Metadata embedded in social content supplements explicit data, including geolocation tags, device identifiers, IP addresses, and timestamps, which facilitate spatiotemporal profiling without relying on overt declarations. For example, location data from check-ins or geotagged posts has been used to correlate mobility patterns with socioeconomic traits in studies analyzing over 10 million Twitter users. Publicly available datasets, such as those from Reddit or YouTube APIs, extend sources beyond proprietary platforms, providing anonymized corpora for behavioral modeling, though they often exclude private interactions.^[34]^[35]^[36] Third-party aggregators and data brokers compile social data from multiple networks, enriching it with cross-platform linkages via identifiers like email or phone numbers, though accuracy varies due to consent-based collection limits. Empirical analyses, such as those employing over 500,000 profiles from OSNs, demonstrate that combining textual content with network metrics achieves up to 85% accuracy in trait prediction, underscoring the potency of these sources despite noise from bots or misinformation. Limitations persist, as much data remains siloed by platform policies—e.g., post-2018 Cambridge Analytica fallout prompted API restrictions reducing access volumes by orders of magnitude.^[37]^[38]

Algorithms and Inference Methods

Social profiling relies on algorithms that process heterogeneous data—such as text posts, network connections, and interaction patterns—to infer attributes like demographics, interests, and personality traits. Machine learning techniques predominate, including supervised classifiers trained on labeled datasets to predict binary or categorical attributes; for instance, support vector machines (SVM) and random forests have been applied to infer gender and age from linguistic features in social media profiles with accuracies exceeding 80% in controlled studies.^[39] Content-based inference methods extract features from user-generated text using natural language processing (NLP) tools, such as bag-of-words representations, TF-IDF weighting, or transformer-based embeddings from models like BERT, to model topical interests or sentiments. Topic modeling algorithms, notably Latent Dirichlet Allocation (LDA), decompose post corpora into latent themes, enabling inference of user preferences; empirical evaluations on platforms like Twitter demonstrate LDA's utility in identifying political leanings by aligning user topics with known ideological clusters.^[9] ^[40] In contrast, relational or network-based methods leverage graph algorithms under the homophily principle, where similar users connect preferentially; label propagation and collective classification propagate known attributes across egonets, inferring missing traits like occupation from friends' profiles with reported F1-scores up to 0.75 on datasets from Sina Weibo.^[41] ^[42] Hybrid approaches integrate multiple modalities, employing ensemble models or multi-level inference frameworks that fuse content signals with behavioral logs (e.g., posting frequency) and structural features (e.g., centrality measures). For example, semi-supervised learning variants, such as collective semi-supervised methods, bootstrap inferences from sparse labels using convex optimization over graph Laplacians, achieving robust performance on incomplete social graphs.^[41] Clustering algorithms like k-means or spectral clustering group users into profiles based on aggregated features, facilitating scalable inference without exhaustive labeling; these have been validated on large-scale networks for tasks like interest prediction, though they risk amplifying biases in training data distributions.^[9] Deep learning extensions, including graph neural networks (GNNs), embed relational data for end-to-end attribute prediction, outperforming traditional methods on dynamic platforms by capturing temporal evolutions in user behavior.^[31] Inference accuracy varies by attribute and platform, with demographic traits like gender often more reliably predicted (AUC > 0.9 via stylistic cues) than nuanced ones like personality, where Big Five traits inference from text yields correlations around 0.3-0.5 with ground truth surveys. These methods assume causal links between observable signals and latent traits, grounded in empirical patterns rather than pure correlation, but require validation against holdout data to mitigate overfitting.^[31]

Integration with AI and Machine Learning

Machine learning algorithms enhance social profiling by automating the extraction of inferences from large-scale social data, such as user posts, connections, and interactions, enabling predictions of traits like demographics, interests, and behaviors that surpass manual analysis in speed and scale.^[31] Techniques including supervised learning classify user attributes based on labeled datasets, while unsupervised methods like clustering group similar profiles from aggregated behavioral data.^[9] For instance, natural language processing (NLP) models analyze textual content to infer sentiment and topics, with studies demonstrating accuracy rates exceeding 85% in personality trait prediction from social media posts.^[43] Deep learning architectures, such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs), further refine profiling by learning hierarchical features from multimodal data like text and images.^[44] Graph neural networks process social connections to model relational dynamics, improving inferences on influence and community affiliations, as evidenced by applications achieving up to 95% precision in influencer identification via PageRank-integrated models.^[45] These methods dynamically update profiles by tracking evolving user activities and reformulating them against topic categories, with similarity metrics guiding real-time adaptations.^[38] Integration challenges include algorithmic biases inherited from training data, often reflecting skewed representations in social platforms, which can amplify errors in profiling underrepresented groups.^[46] Empirical validations, such as those using long short-term memory (LSTM) networks for behavioral pattern recognition, report F1-scores above 0.90 in user categorization tasks, underscoring causal links between data quality and predictive reliability.^[47] Recent advancements, including transformer-based models like those in GPT variants, enable zero-shot learning for nuanced inferences, though their deployment requires scrutiny of source data credibility to mitigate overgeneralization.^[48]

Commercial Applications

Marketing and Targeted Advertising

Social profiling in marketing involves aggregating data from individuals' online social interactions—such as likes, shares, comments, and network connections—to infer demographic, psychographic, and behavioral attributes, enabling advertisers to deliver tailored content. Platforms like Facebook and Google utilize these profiles to segment users beyond basic demographics, incorporating inferred interests and personality traits derived from social graph data. For instance, analysis of a user's liked pages or followed influencers can predict preferences for products ranging from consumer goods to financial services, facilitating micro-targeting that aligns ad creatives with specific motivations.^[49]^[50] Empirical studies demonstrate the efficacy of such profiling in enhancing ad performance. Psychological targeting, which matches ad appeals to inferred personality traits from social media activity, has been shown to increase click-through rates by up to 40% and positive message responses by a factor of 1.23 compared to non-targeted approaches in controlled experiments involving over 3.5 million Facebook users. Similarly, trait-based personalization in social ads yields higher engagement when aligned with users' openness or extraversion scores, though effects vary by trait and product type, with meta-analyses indicating average uplift in persuasiveness of 10-20% over generic ads. These outcomes stem from causal mechanisms where profiled relevance reduces ad fatigue and boosts perceived utility, as evidenced by randomized trials controlling for confounders like ad frequency.^[49]^[51]^[52] In terms of return on investment (ROI), targeted advertising leveraging social data outperforms broad campaigns. As of 2023, global marketers reported Facebook and Instagram—platforms heavily reliant on social profiling—as delivering the highest ROI among social channels, with average returns exceeding those of non-profiled media by factors of 2-3x in conversion metrics. Personalization driven by social inferences has been linked to a 202% increase in conversion rates and 30% higher ROI in AI-augmented systems, based on aggregated data from over 1,000 ad campaigns analyzed in 2024-2025 reports. Ad spend on such platforms reached $276 billion in 2025, with profiling enabling cost efficiencies by prioritizing high-propensity audiences, reducing cost-per-acquisition by 15-25% in sectors like e-commerce and finance.^[53]^[54]^[55] Commercial implementations often integrate social profiles with first-party data for retargeting. Companies like Procter & Gamble and Coca-Cola have employed social network-based targeting to amplify reach, where ads referencing a user's connections (e.g., "Friends like this") increase purchase intent by 10-15%, per field experiments. However, effectiveness depends on data accuracy; studies note that inferred profiles from social signals achieve 70-85% precision in behavioral predictions, outperforming traditional surveys but susceptible to noise from performative online behavior. Regulatory shifts, such as Apple's 2021 App Tracking Transparency framework, have prompted adaptations like contextual profiling, yet core ROI advantages persist in compliant ecosystems.^[50]^[56]

Influence Measurement and Reputation Scoring

Commercial entities employ social profiling to quantify individuals' online influence, primarily for selecting influencers in marketing campaigns. Algorithms analyze aggregated data from platforms like Twitter (now X) and Instagram, including follower counts, post engagement (likes, shares, comments), and network centrality, to generate scores predicting content dissemination potential.^[57] For instance, engagement rate, computed as total interactions divided by followers multiplied by 100, serves as a core metric, with rates above 1-3% indicating higher influence in niche audiences.^[58] Reach, measuring unique users exposed to content, and impressions, counting total views, further refine profiles by inferring audience scale from behavioral patterns.^[59] Historical systems like Klout, operational from 2008 to 2018, exemplified this by producing a proprietary score from 1 to 100 based on users' cross-platform activity, emphasizing retweet frequency, mention velocity, and audience amplification over mere follower volume.^[60] Acquired by Lithium Technologies in 2014, Klout's model weighted true reach—verified engaged followers—against spam signals, though critics noted its opacity and susceptibility to gaming via reciprocal interactions.^[61] Modern successors in influencer marketing, such as Traackr and Upfluence, integrate similar profiling with AI-driven fake follower detection and demographic inferences from likes and shares, enabling brands to score influencers' authentic sway.^[62] These tools often benchmark against industry averages, where top micro-influencers (10,000-100,000 followers) yield engagement rates of 2.5% as of 2024 data.^[63] Reputation scoring via social profiling extends influence metrics to assess perceived trustworthiness and sentiment, crucial for commercial endorsements and partnerships. Systems aggregate social mentions, review platforms, and interaction histories to compute composite scores, factoring negative signals like controversy flags or bot-like behavior.^[64] Earned media value (EMV), valuing organic exposure at paid ad equivalents (e.g., $0.01-0.05 per impression), quantifies reputational leverage, with campaigns averaging $5.78 return per dollar spent in 2023 analyses.^[65] Tools like Reputation.com derive business-level scores from social sentiment and review volumes, applying weighted algorithms where recent data predominates, though individual applications remain niche due to data privacy constraints.^[66] In practice, high-reputation profiles exhibit consistent positive sentiment ratios above 80%, inferred from natural language processing of comments and shares.^[67] Such scoring mitigates risks in collaborations by profiling causal links between past behaviors and future endorsements, prioritizing empirical engagement over self-reported claims.

Employment and Consumer Credit Decisions

In employment decisions, employers increasingly employ social profiling by analyzing candidates' publicly available social media activity to evaluate traits such as professionalism, cultural fit, and risk factors like dishonesty or substance abuse. A 2023 ResumeBuilder survey reported that 73% of hiring managers conduct social media screenings as part of recruitment.^[68] Similarly, 85% of recruiters integrate such checks into hiring processes, often focusing on platforms like LinkedIn, where 87% identify it as the most effective for assessing applicants.^[69]^[70] These practices have tangible impacts, with a 2018 CareerBuilder survey indicating that 70% of employers review profiles and 54% reject candidates based on findings such as inappropriate content or inconsistent information.^[71] Empirical analyses show that problematic social media postings can negatively influence hiring perceptions by signaling potential behavioral issues, though outcomes vary by content type and recency.^[72] For consumer credit decisions, social profiling supplements traditional metrics by leveraging online social data—such as network connections, posting patterns, and digital footprints—to assess creditworthiness, especially for underbanked populations lacking conventional histories. Fintech firms like ZestFinance and Kreditech use these data in lending models, where website access and social activity predict default rates more accurately than legacy scores alone, as demonstrated in FDIC analyses of digital footprint efficacy.^[73] Academic studies confirm that incorporating social network data enhances score precision; for instance, borrowers' connections and activity correlate with repayment behavior, enabling strategic network adjustments to improve assessments.^[74] Recent machine learning applications further refine this by analyzing sentiment and engagement from social media to forecast individual risk, outperforming baseline models in empirical tests on loan portfolios.^[75] Such methods expand access in emerging markets but raise questions about data granularity's causal link to financial reliability, with evidence suggesting modest predictive gains over socioeconomic proxies.^[76]

Public and Governmental Uses

Law Enforcement and Security Profiling

Social profiling in law enforcement and security contexts involves the systematic analysis of publicly available online data, including social media posts, connections, and behavioral patterns, to assess risks of criminal activity, terrorism, or threats to public safety. Agencies such as the Federal Bureau of Investigation (FBI) and Department of Homeland Security (DHS) employ these methods to generate investigative leads, monitor known suspects, and screen individuals for potential dangers. For instance, the FBI integrates social media intelligence into counterterrorism operations, where agents review posts for indicators of radicalization or planning, often yielding actionable tips from public sources.^[77] Local police departments, like the Philadelphia Police Department, have utilized platforms such as Twitter and Facebook since at least 2013 to track emerging threats in real time, contributing to arrests in cases involving gang activity or public disturbances.^[77] In immigration and border security, the U.S. government mandates disclosure of social media identifiers for visa applicants, a policy piloted in 2014 and expanded to all visa waiver program entrants by 2019, with further enhancements in 2025 targeting antisemitic activity and student visas.^[78] This screening has enabled DHS to identify discrepancies between applicants' online expressions and stated intentions, flagging associations with extremist groups or support for violence, though official reports emphasize its role in layered vetting without disclosing specific denial statistics.^[79] Predictive applications extend to algorithms that incorporate social data for hotspot forecasting or individual risk scoring, as seen in some municipal programs, where historical online patterns inform patrol allocations.^[80] Empirical assessments of effectiveness remain constrained by classified operations and methodological challenges, but surveys of over 500 U.S. law enforcement agencies indicate that 81% actively use social media for investigations, with 73% reporting it aids in solving crimes and 67% in anticipating them through early detection of threats like planned violence announced online.^[81] Case examples include disruptions of plots where social posts revealed weapon acquisitions or recruitment efforts, aligning with causal links between digital signaling and offline actions in terrorism contexts.^[82] While rigorous randomized studies are scarce, practitioner data suggest these tools enhance response times and resource efficiency, particularly for transient threats not captured by traditional records.^[83] China's Social Credit System (SCS), launched through a 2014 State Council planning outline, represents the most comprehensive national effort to integrate social profiling into governance, aiming to foster trustworthiness across individuals, businesses, and government entities by 2020. Although a unified nationwide scoring mechanism remains absent as of 2023, the system encompasses fragmented pilots, blacklists, and data-sharing platforms that track compliance with laws, financial obligations, and social norms, drawing from over 80 local and sectoral initiatives.^[84] Official goals emphasize rectifying "dishonest" behaviors, such as court judgment evasion or corporate regulatory violations, with data sourced from judicial records, tax filings, and public behavior like traffic infractions.^[85] Implementation varies by region and sector, with corporate-focused elements under the Corporate Social Credit System (CSCS) achieving broader rollout since 2014, including public ratings for over 33 million firms by 2020 based on metrics like contract fulfillment and environmental compliance.^[86] For individuals, joint incentive and punishment mechanisms—effective from 2018—have led to tangible restrictions: by 2022, approximately 8.5 million "dishonest" persons were barred from high-speed rail travel, and millions faced loan denials or employment limits in state sectors.^[87] Empirical analyses indicate these measures correlate with behavioral shifts, such as a 10-15% drop in court judgment non-compliance post-pilot launches in cities like Rongcheng, where localized scoring influenced 1.6 million residents' access to services.^[88] However, benefits skew toward politically aligned entities, with connected firms gaining higher scores via sanctioned donations, raising questions of selective enforcement.^[88] Beyond punitive tools, the SCS promotes positive incentives, such as whitelisting compliant entities for expedited approvals; a 2023 regulation enabled qualifying companies to bypass certain customs inspections.^[89] Recent action plans, like the 2024-2025 National Development and Reform Commission outline, prioritize data interoperability to address "information silos," integrating AI for real-time monitoring while expanding to supply chain trustworthiness.^[85] Studies link the system to enhanced firm investment efficiency, reducing agency costs and overinvestment by up to 5-7% in pilot areas through improved information transparency.^[90]^[91] No equivalent full-scale national systems exist elsewhere, though proposals in Russia and data governance frameworks in India (e.g., Aadhaar) share superficial parallels in citizen tracking without formalized social penalties.^[92] Foreign businesses operating in China face indirect extraterritorial pressures, as non-compliance can propagate via global supply chains.^[93]

Benefits and Empirical Outcomes

Efficiency and Personalization Advantages

Social profiling facilitates efficient resource allocation in targeted advertising by analyzing users' online behaviors, networks, and expressed interests to prioritize high-potential audiences, thereby reducing ad waste and boosting return on investment. Empirical analyses indicate that social media marketing strategies leveraging such data can enhance sales through improved attention and engagement, with sequential targeting models demonstrating superior performance over broad campaigns.^[94] Personalization efforts informed by social data have been associated with revenue lifts of 10 to 15 percent across digital channels, as they minimize irrelevant exposures and maximize relevance.^[95] In consumer credit assessment, social profiling augments traditional models with network and behavioral indicators, improving predictive accuracy for default risk, particularly for individuals lacking conventional credit histories. Research utilizing social network data has shown enhancements in model performance metrics, such as area under the curve (AUC) scores, by incorporating relational ties and activity patterns that signal financial reliability.^[96] This efficiency stems from real-time data integration, enabling faster evaluations and lower operational costs compared to manual underwriting expansions.^[97] Similarly, in employment screening, profiling via social media yields insights into cultural fit and behavioral red flags, correlating with reduced turnover rates by aiding selections that align with organizational needs.^[98] Personalization advantages manifest in heightened user engagement and satisfaction on social platforms, where algorithmically tailored content feeds—derived from profiling—increase metrics like time spent, shares, and interactions. Studies on AI-driven personalization reveal positive effects on customer experience, with users reporting greater perceived value, enjoyment, and convenience from relevant recommendations.^[99] Approximately two-thirds of consumers have encountered these benefits, underscoring how profiling counters information overload by surfacing causally aligned content, thereby sustaining platform utility without excessive manual curation.^[100]

Evidence from Security and Economic Impacts

In security applications, social profiling via social network analysis has facilitated the disruption of criminal and terrorist organizations by pinpointing central actors whose removal degrades network efficacy. Empirical applications in predictive policing, which leverage behavioral and relational data patterns, have yielded measurable crime reductions; for example, in Santa Cruz, California, algorithm-driven deployment targeting burglary hotspots resulted in a 19% decline in burglaries over six months without additional personnel.^[101] In a controlled Los Angeles Police Department experiment, similar profiling techniques doubled forecasting accuracy and produced a 12% drop in property crimes in treated areas, contrasting with a 0.4% citywide rise.^[101] Broader integrations of AI and network data suggest potential urban crime reductions of 30-40%, as estimated by analyses of proactive law enforcement strategies.^[102] Economically, social profiling through digital footprints and online behavioral data enhances credit risk models, improving default prediction and enabling lending to previously unscorable individuals. A study of over 200,000 German loan applicants found that digital footprint variables alone achieved an area under the curve (AUC) of 69.6% for default forecasting, surpassing traditional credit bureau scores (68.3%), with a combined model reaching 73.6%—a 5.3 percentage point gain—while correlating minimally (10%) with bureau data to provide complementary insights.^[73] This approach supports financial inclusion for approximately 2 billion unbanked adults globally by extending credit access without proportionally elevating risks.^[73] In underserved populations exceeding one million, AI models incorporating alternative social data increased loan approvals and decreased default rates simultaneously, yielding net economic gains through expanded consumption and GDP contributions in emerging markets.^[103] For fraud detection, social media pattern analysis complements conventional methods, identifying anomalous behaviors that mitigate financial losses, though quantified savings vary by implementation.^[104]

Counterarguments to Overstated Harms

Critics of social profiling frequently emphasize risks such as privacy erosion and discriminatory outcomes, yet empirical evidence suggests these harms are often exaggerated relative to demonstrated benefits. The "privacy paradox" illustrates this discrepancy: surveys indicate high stated concerns about data usage, but behavioral data reveals widespread voluntary disclosure of personal information on social platforms, with users prioritizing utility like connectivity and personalization over abstract privacy fears.^[105]^[106] For instance, longitudinal analyses confirm that privacy attitudes do not consistently predict reduced sharing, implying that perceived harms do not substantially deter participation and that regulatory responses may overstate real-world impacts.^[107] Algorithmic profiling, by leveraging large datasets, can mitigate rather than amplify biases inherent in human decision-making. Unlike subjective judgments prone to fatigue, stereotypes, or inconsistencies, data-driven models apply consistent rules across populations, reducing variability in outcomes.^[108]^[109] In employment and lending contexts, studies show social media-derived profiles enhance predictive accuracy without introducing novel discriminatory effects beyond those in traditional methods, as algorithms lack intrinsic preferences for group-based harm.^[110] Specific applications underscore net positive empirical outcomes. In credit scoring, incorporating social network data has improved model precision, enabling higher approval rates for underserved borrowers with limited financial histories while lowering default losses for lenders—evidenced by peer-to-peer platforms reporting reduced risk exposure post-implementation.^[96]^[111] Similarly, predictive policing augmented by social media signals has demonstrated crime reductions exceeding 7-20% in targeted areas compared to conventional patrols, with false positive rates not significantly diverging from human-led predictions when calibrated properly.^[101]^[112] These gains suggest that while risks exist, they are often contextually manageable and outweighed by efficiency improvements, challenging narratives of pervasive, unmitigated harm.

Criticisms and Ethical Challenges

Privacy Invasions and Data Leaks

Social profiling frequently entails the aggregation of personal data from social media platforms, online behaviors, and public records without explicit user consent, constituting a direct invasion of privacy. In the 2018 Cambridge Analytica scandal, the firm illicitly obtained data on over 50 million Facebook users through a third-party app, enabling psychographic profiling for targeted political advertising.^[28] This unauthorized harvesting demonstrated how profiling systems exploit platform APIs and user networks to infer sensitive traits like personality and political leanings, often bypassing privacy settings.^[28] Facial recognition technologies integrated with social profiling exacerbate these invasions by scraping billions of publicly posted images. Clearview AI compiled a database of more than 30 billion facial images primarily sourced from social media sites without individuals' knowledge or permission, facilitating real-time identification and behavioral inference.^[113] Regulatory bodies have imposed substantial fines for such practices; for instance, the Dutch Data Protection Authority levied €30.5 million on Clearview AI in September 2024 for violating privacy laws through mass data collection.^[113] Similarly, France's CNIL fined the company €20 million in 2021 for unauthorized image scraping from French residents.^[114] Data brokers further amplify risks by purchasing and merging social media-derived datasets into comprehensive consumer profiles sold to marketers and employers, often without transparency or opt-out mechanisms.^[115] Data leaks compound these vulnerabilities, exposing aggregated profiles to unauthorized access and exploitation. In 2019, a major data broker inadvertently exposed a database containing 235 million social media profiles, including emails, phone numbers, and locations, which could be repurposed for advanced profiling or identity theft.^[116] More recently, the 2024 National Public Data breach compromised over 2.9 billion personal records from a data broker specializing in background checks and profiling, revealing sensitive identifiers like Social Security numbers and addresses derived partly from social sources.^[117] Such incidents enable malicious actors to enrich stolen data with social profiling insights, heightening risks of targeted scams and surveillance.^[118] Government social media monitoring programs also face leak risks; U.S. federal agencies' vetting of immigrants' online activity has led to documented privacy erosions and potential data exposures through inadequate safeguards.^[119] In systems like China's social credit framework, centralized data aggregation for behavioral scoring inherently elevates leak probabilities due to the scale of surveillance involved, though specific breaches remain underreported amid state controls.^[120] Overall, these events underscore systemic frailties in profiling infrastructures, where vast datasets invite both intentional invasions and accidental disclosures, often prioritizing utility over robust privacy protections.^[121]

Algorithmic Bias and Discriminatory Effects

In social profiling applications, algorithmic bias manifests when predictive models, trained on historical or behavioral data, produce outcomes that disproportionately disadvantage certain demographic groups, often through proxies correlated with protected attributes like race or ethnicity. Such biases can arise from skewed training datasets reflecting past discriminatory practices or unrepresentative sampling, leading to amplified errors in areas like risk assessment or ad targeting. However, empirical analyses reveal that many disparities labeled as bias actually mirror underlying base rate differences in behaviors or outcomes across groups, rather than flawed model design.^[122] A prominent example is the COMPAS recidivism prediction tool, deployed in U.S. courts to inform sentencing and parole since the 2010s. A 2016 ProPublica investigation of over 7,000 Florida defendants found Black individuals scored as high-risk were twice as likely as whites to be false positives—falsely predicted to reoffend (45% vs. 23%)—while whites had higher false negative rates.^[123] This prompted widespread criticism of racial bias. Yet subsequent critiques, including reanalyses of the same data, showed COMPAS achieved predictive parity (calibration) across races, with similar positive predictive values for both groups around 60-65%. The error rate imbalances stem from higher base recidivism rates among Black defendants (approximately 60% vs. 40% for whites in the dataset), rendering equal false positive/negative rates mathematically impossible without reducing overall accuracy, as confirmed by impossibility theorems in algorithmic fairness literature.^[124]^[125] Forcing parity would misclassify more individuals overall, potentially increasing societal risks from released offenders. Similar patterns appear in facial recognition systems used for social surveillance and profiling, where early studies reported error rates up to 34 times higher for darker-skinned females compared to light-skinned males. U.S. National Institute of Standards and Technology (NIST) evaluations from 2019, testing 189 algorithms on millions of images, attributed much of this to demographic differentials in image quality and acquisition conditions, not algorithmic racism; top-performing commercial systems exhibited near-zero bias, with false positive disparities dropping below 0.1% when controlling for confounders.^[126] In social media profiling for hiring or advertising, algorithms infer traits from online activity, but uneven platform usage—e.g., lower representation of certain minorities in training data—can yield disparate ad delivery, as seen in 2019 studies of STEM job ads shown less to women due to behavioral proxies like past clicks, though effects diminished with diverse data auditing.^[127] Discriminatory effects, such as denied opportunities or heightened scrutiny, occur when these systems inform high-stakes decisions, potentially perpetuating cycles of exclusion. In lending and employment profiling using social data, opaque models have been linked to higher rejection rates for protected groups, but controlled experiments indicate algorithms often exhibit less bias than human decision-makers, who introduce subjective heuristics; for instance, a 2020 review found AI-driven credit scoring reduced racial disparities by 40% compared to traditional methods reliant on manual review. Critics from advocacy groups emphasize disparate impact under legal standards like the U.S. Equal Credit Opportunity Act, yet causal analyses underscore that interventions prioritizing group equity over predictive validity—e.g., via fairness constraints—can degrade model performance by 10-20%, harming neutral users and efficiency.^[122] Empirical evidence thus suggests that while genuine technical biases warrant mitigation through diverse datasets and audits, overstated claims of discrimination frequently overlook how accurate profiling captures real, empirically observed group variances in outcomes like recidivism or creditworthiness, which stem from socioeconomic and behavioral factors rather than algorithmic malice.^[128]

Regulatory Overreach and Free Speech Implications

Regulations mandating social profiling on digital platforms to enforce content moderation standards have raised concerns over governmental overreach, as they compel private entities to surveil and suppress user expression preemptively. The European Union's Digital Services Act (DSA), effective for very large online platforms since August 2024, requires operators to conduct systemic risk assessments that involve analyzing user-generated content and behavioral patterns to identify and mitigate dissemination of illegal or harmful material, such as disinformation or hate speech, with noncompliance penalties reaching 6% of global annual turnover. Critics, including legal scholars, argue this framework incentivizes platforms to err toward excessive removal of borderline content to minimize liability, effectively outsourcing censorship to algorithms and human moderators who profile speech based on vague risk criteria, thereby chilling protected expression without due process.^[129]^[130] Such mandates extend implications beyond the EU due to the extraterritorial "Brussels effect," where platforms apply uniform global policies to avoid fragmented compliance, potentially restricting lawful speech in jurisdictions with stronger free speech protections, like the United States. For instance, the DSA's obligations for transparency in content moderation decisions and algorithmic profiling have prompted platforms to enhance surveillance of user profiles for systemic risks, leading to documented increases in proactive content flagging and removal rates post-enforcement; a 2025 analysis highlighted opaque data practices that obscure whether these measures disproportionately target dissenting viewpoints on topics like elections or public health. Organizations defending free expression contend this regulatory pressure transforms platforms into de facto state agents, eroding the voluntary nature of moderation and infringing on Article 10 of the European Convention on Human Rights, which safeguards even provocative speech absent imminent harm.^[131]^[132]^[133] In the United States, analogous overreach manifests through proposed reforms to Section 230 of the Communications Decency Act of 1996, which shields platforms from liability for user content but faces scrutiny for enabling unchecked moderation; bills like the 2023 EARN IT Act and others condition immunity on "reasonable" efforts to combat certain profiled behaviors, such as child exploitation material, but extend to broader categories like misinformation, prompting fears of compelled profiling that amplifies self-censorship. The Supreme Court's June 2024 ruling in Murthy v. Missouri addressed federal agencies' communications with platforms urging removal of COVID-19-related posts, dismissing the case on standing grounds but underscoring potential First Amendment violations if such jawboning coerces speech restrictions based on behavioral profiling. Empirical evidence from platform disclosures shows a spike in content takedowns following government inquiries—e.g., over 10 million posts flagged in 2021-2022 for pandemic skepticism—illustrating how regulatory threats distort moderation incentives away from neutral facilitation toward viewpoint-discriminatory enforcement.^[134]^[135]^[136] These developments highlight a causal tension: while aimed at curbing verifiable harms like incitement, overbroad profiling mandates undermine causal accountability by prioritizing regulatory compliance over evidence-based harm thresholds, fostering environments where empirical dissent on policy issues faces heightened scrutiny. Policy analyses from free speech advocates emphasize that without precise, judicially reviewable criteria, such systems risk entrenching biases in enforcement—e.g., disproportionate targeting of conservative-leaning profiles in moderation logs—exacerbating distrust in institutions already perceived as ideologically skewed.^[137]^[138]

Future Implications

Technological Evolutions and Innovations

Advancements in machine learning have enabled social profiling to evolve from rudimentary statistical correlations to sophisticated predictive models that infer traits from online interactions. Since the early 2010s, supervised learning techniques, such as support vector machines and random forests, have been supplanted by deep neural networks, achieving higher precision in identifying user intents and preferences from social media data.^[139] A 2024 comprehensive survey documents this progression, noting the integration of graph neural networks for analyzing social connections, which improved community detection and influence prediction by up to 20% in benchmark datasets compared to traditional methods.^[31] Natural language processing innovations, particularly transformer-based architectures like BERT introduced in 2018, have enhanced trait inference from textual content, enabling models to predict psychological attributes such as extroversion or risk propensity with accuracies around 65-80% across validation sets drawn from platforms like Twitter and Facebook.^[31] Concurrently, computer vision advancements allow profiling via image and video analysis, where convolutional neural networks extract behavioral cues from user-generated media, correlating visual styles with socioeconomic indicators.^[45] Multimodal AI systems represent a key 2020s innovation, fusing text, visuals, and metadata for comprehensive user representations; a 2024 review underscores their role in behavioral analysis, where foundation models process diverse inputs to forecast actions with reduced error rates.^[140] In predictive applications, large language models with browsing capabilities, as demonstrated in mid-2025 experiments, autonomously retrieve public social profiles to infer demographics like age and location, attaining prediction accuracies of 70-85% on held-out samples without fine-tuning.^[141] Emerging federated learning frameworks address data silos in profiling, training models across decentralized devices while preserving privacy, with pilot implementations showing comparable performance to centralized approaches in user segmentation tasks since 2023.^[31] For real-time systems akin to national scoring mechanisms, AI-driven facial recognition has scaled behavioral monitoring, as integrated into China's social credit infrastructure by 2024, processing millions of daily identifications to update scores dynamically.^[87] These evolutions prioritize computational efficiency, with edge AI deployments reducing latency to milliseconds for on-device profiling.^[31]

Balancing Utility with Accountability

Regulatory frameworks such as the European Union's Artificial Intelligence Act, which entered into force on August 1, 2024, and imposed prohibitions on unacceptable-risk practices including certain forms of social scoring by February 2, 2025, classify biometric categorization and emotion recognition in public spaces as high-risk, mandating risk management systems, data governance, and human oversight to ensure accountability while permitting utility in vetted applications like law enforcement.^[142]^[143] High-risk profiling systems under the Act require conformity assessments, transparency reporting, and post-market monitoring, with fines up to €35 million or 7% of global turnover for non-compliance, aiming to mitigate discriminatory effects without fully curtailing predictive benefits observed in sectors like fraud detection where algorithms have reduced false positives by up to 20% in controlled pilots.^[144]^[145] In the United States, the National Institute of Standards and Technology's 2021 framework for automated decision-making emphasizes accountability through impact assessments, bias mitigation, and audit trails, applied to social profiling in contexts like predictive policing, where tools analyzing social media patterns have forecasted crime hotspots with 5-10% accuracy gains over traditional methods but faced scrutiny for lacking independent validation.^[146] Empirical evaluations of such systems, including a 2020 analysis of location-based predictive policing, indicate that incorporating accountability measures like external audits correlates with reduced error rates in high-stakes deployments, though opaque algorithms persist as a barrier, with only 15% of surveyed U.S. agencies reporting routine third-party reviews as of 2023.^[147]^[148] Technical approaches like differential privacy enable balancing utility and privacy by adding calibrated noise to datasets, preserving analytical value for profiling tasks; a 2024 clinical study demonstrated that k-anonymity de-identification retained 85-95% of predictive utility for risk modeling while complying with privacy thresholds, outperforming stronger anonymization that degraded accuracy by over 30%.^[149] In social media surveillance for security, Deloitte's 2023 review of urban AI implementations found that hybrid models combining algorithmic predictions with human review achieved 70% efficacy in threat detection while adhering to data minimization principles, underscoring causal trade-offs where excessive accountability layers can increase operational costs by 25% but avert overreach liabilities.^[150] Challenges remain, as mixed empirical outcomes from accountability initiatives—such as variable impacts on bias reduction in profiling datasets—highlight the need for ongoing validation, with peer-reviewed meta-analyses confirming no universal formula but context-specific calibration as essential for causal efficacy.^[151]

References

[1]
Social Profiling: A Review, Taxonomy, and Challenges
The two major categories that exist in social profiling are individual and group profiling. We present an overview of these categories in this section. Figure 3 ...
[2]
What is social media profiling? | Eftsure US
Social profiling also furthers demographic and psychographic segmentation to develop hyper-targeted marketing campaigns that address customer needs and resonate ...
[3]
Securing social spaces: machine learning techniques for fake profile ...
Dec 8, 2024 · Studies indicate that machine learning can effectively detect fraudulent profiles, typically assessed using accuracy, F1-score, and recall ...
[4]
Tools, Potential, and Pitfalls of Social Media Screening
Sep 18, 2023 · Social media screening, or social profiling, is an emerging approach to assessing candidates' social influence, personalities, and workplace behaviors.
[5]
Personality prediction model for social media using machine ...
A maximum F1-score of 0.762 and accuracy of 78.34% on the Facebook dataset; 0.783 and 79.67%; on the Twitter dataset, 0.821; 86.84% on Instagram dataset is ...
[6]
Fake Social Media Profile Detection Using Machine Learning and ...
The accuracy for detection of fake profiles through these means is 96.5%. This offers a complete method for detection of fake profiles through the synergy of ...
[7]
Stereotype Accuracy is One of the Largest and Most Replicable ...
Feb 16, 2016 · Social psychologists have many “basic phenomena” that are presumed (without evidence) to cause inaccuracy: categorization supposedly exaggerates ...
[8]
Why accuracy dominates bias and self-fulfilling prophecy
Jun 16, 2015 · However, accuracy itself is controversial in social psychology, and those controversies (Section 6: “Accuracy controversies”) and some key ...
[9]
The rise of user profiling in social media: review, challenges and ...
Oct 19, 2023 · Social network analysis (SNA) is a popular technique associated with big data analytics for user profiling in social media to discover patterns, ...
[10]
Social Profiling: A Review, Taxonomy, and Challenges | Request PDF
This study reviews and classifies research inferring users social profile attributes from social media data as individual and group profiling.Missing: evidence | Show results with:evidence
[11]
https://nordvpn.com/cybersecurity/glossary/social-profiling/
[12]
What Is Social Media Profiling - How It Works? - SEON
Social media profiling refers to using information that is publicly available on social media platforms to create someone's profile.Missing: distinctions psychological
[13]
Social Profiling: A Review, Taxonomy, and Challenges - PubMed
This study reviews and classifies research inferring users social profile attributes from social media data as individual and group profiling.Missing: empirical | Show results with:empirical
[14]
What Is Data Profiling? Ensuring Data Trust - BigID
May 24, 2024 · Data profiling is the systematic analysis of data to uncover its structure, quality, and content. It involves examining various attributes within datasets.
[15]
Psychological profiling | Research Starters - EBSCO
Psychological profiling is a practice used primarily in criminal investigations to infer the characteristics of perpetrators based on the details of the crimes ...
[16]
Psychological Profiling - Sage Publishing
Sep 18, 2012 · As we use the term here, however, psychological profiling refers to gathering data on a known individual or individuals, such as by background ...
[17]
The Early Days of Web Analytics - Amplitude
Early 1900s to 1960: Researchers begin experimenting with wireless data transmission. · 1990: The internet is born. · 1993: The first rudimentary web-analytics ...<|separator|>
[18]
A History of the Data-Tracked User | The MIT Press Reader
Oct 8, 2021 · A brief history of the commercially targeted user, beginning with “identity scoring” in the 1940s and ending with the targeted advertising of today.
[19]
25 years of digital analytics with Brian Clifton - Piwik PRO
Apr 24, 2024 · The introduction of cookies by Netscape in 1994 allowed for more sophisticated tracking of user behavior, providing advertisers with valuable ...
[20]
03. The History of Digital Advertising Technology - AdTech Book
DoubleClick was founded in 1995 by Kevin O'Connor and Dwight Merriman. It was purchased by private equity firms Hellman & Friedman and JMI Equity in July 2005 ...
[21]
Tracking the trackers: Introduction to cookies and web tracking
Apr 23, 2012 · Web beacons: these are small objects embedded into a web page, but are not visible. They can also be known as "tags", "tracking bugs", "pixel ...
[22]
The rise of social media - Our World in Data
Sep 18, 2019 · Social media started in the early 2000s. MySpace was the first social media site to reach a million monthly active users – it achieved this ...
[23]
The Evolution of Social Media: How Did It Begin, and Where Could It ...
In a sense, social media began on May 24, 1844, with a series of electronic dots and dashes tapped out by hand on a telegraph machine. The first electronic ...
[24]
(PDF) The rise of user profiling in social media: review, challenges ...
... social proﬁling has. surfaced as a viable approach by incorporating a personal-. ized search while maintaining the exclusiveness of the user. proﬁles ...
[25]
History of Social Media (It's Younger Than You Think)
Sep 19, 2025 · But the first real social network appeared in 1997 with SixDegrees, a platform that allowed profiles and connections. Blogs and early networks ...
[26]
A Brief History of Online Advertising - HubSpot Blog
Nov 29, 2021 · 2006: Digital ads become hyper-targeted. As social media platforms picked up steam in the mid 2000s, advertisers sought a way to integrate ad ...
[27]
Targeted Advertising: Origins, Methods, and Future Trends
May 13, 2024 · The origins of targeted advertising can be traced back to the early days of print media. ... social media platforms. This gave advertisers access ...What is targeted advertising? · What is the main goal of... · What types of targeted...
[28]
Revealed: 50 million Facebook profiles harvested for Cambridge ...
Mar 17, 2018 · Whistleblower describes how firm linked to former Trump adviser Steve Bannon compiled user data to target American voters.
[29]
The scant science behind Cambridge Analytica's controversial ...
Mar 29, 2018 · Nature peers into the evidence for 'psychographic targeting'. ... The practices of Cambridge Analytica, a data-analytics firm involved in US ...
[30]
Cambridge Analytica's black box - Margaret Hu, 2020 - Sage Journals
Aug 24, 2020 · The Cambridge Analytica–Facebook scandal led to widespread concern over the methods deployed by Cambridge Analytica to target voters through psychographic ...
[31]
User Modeling and User Profiling: A Comprehensive Survey - arXiv
Feb 20, 2024 · The article outlines a range of user profiling techniques, such as implicit profiling, perceptual preference questionnaires, social profiling ...
[32]
2025 Digital Media Trends | Deloitte Insights
Mar 25, 2025 · Social platforms are extending generative AI tools to help creators run their businesses, create content, target audiences and advertisers, and ...
[33]
Public Data Sources - Media - Madrona Venture Labs
1. Social Media Data: Data from various social media platforms, including posts, comments, likes, shares, followers, and user profiles. This data can provide ...
[34]
Social Data: What It Is, How It Works, and Limitations - Investopedia
Social data is information that users publicly share and includes metadata such as location and biographical data. · Social media companies collect social data ...
[35]
Classifying social position with social media behavioral data
Aug 15, 2025 · This study uses a donation-based data collection, which collected survey responses and digital behavioral data from the same participants. In a ...
[36]
Top 22 Social Media Datasets That You Can Use (Updated in 2025)
Explore top social media data sets to enhance your portfolio. Impress recruiters with our expertly curated list.Meta Datasets And Project... · Reddit Datasets And Project... · Youtube Datasets And Project...
[37]
A Review of User Profiling Based on Social Networks - IEEE Xplore
Sep 11, 2024 · Section IV describes the primary processing phases: data collection, preprocessing, feature extraction, modeling technologies, and performance.
[38]
A model for generating a user dynamic profile on social media
This paper introduces a model for building a dynamic user profile based on the daily activities of the user and other users' activities who are followed by ...2.2. Profiling Osn Users · 5. Experiments And Results · 5.2. Dynamic Profile...
[39]
New machine learning algorithm can predict age and gender from ...
May 16, 2019 · A new “demographic inference” tool developed by academics can make predictions based solely on the information in a person's social media profile.
[40]
[PDF] MLI: A Multi-level Inference Mechanism for User Attributes in Social ...
Content-based methods infer users' attributes based on their text content in the social network such as posted blogs [4, 24], tweets [5, 8, 28, 45], and ...
[41]
[PDF] Collective Semi-Supervised Learning for User Profiling in Social ...
Jun 24, 2016 · Index Terms—Convex optimization, collective learning, semi-supervised learning, social media, user profiling. ... Algorithm. Thanks to the ...
[42]
https://ieeexplore.ieee.org/document/8778611
[43]
AI-Based Psychological Profiling for Social Network Analysis - ESPY
Machine learning algorithms analyze large datasets to identify patterns and make predictions. NLP techniques process and interpret human language from text data ...<|separator|>
[44]
Social network analysis using deep learning: applications and ...
Oct 25, 2021 · Deep learning techniques possess a potential of automatically learning the features of data generated by an OSN to provide appropriate and ...
[45]
Data Profiling and Machine Learning to Identify Influencers from ...
This research aims to find the best machine learning model, using PageRank to profile influential nodes, to identify social media influencers.
[46]
[PDF] Machine Learning Algorithms on Social Media: Privacy Risks, User ...
Oct 4, 2025 · The use of ML to optimize marketing campaigns often involves extensive collection and processing of personal and behavioral data, which can ...
[47]
Countering Social Media Cybercrime Using Deep Learning - MDPI
Oct 11, 2024 · This paper proposes a framework to classify fake and real accounts on Instagram based on a deep learning approach called the Long Short-Term Memory (LSTM) ...<|separator|>
[48]
Advanced machine learning techniques for social support detection ...
This research employs advanced techniques such as Transformers, and zero-shot learning approaches with models like GPT-3, GPT-4, and GPT-4-o. The aim is to ...
[49]
Psychological targeting as an effective approach to digital mass ...
Nov 13, 2017 · Our experiments demonstrate that targeting people with persuasive appeals tailored to their psychological profiles can be used to influence ...
[50]
[PDF] Social Advertising - Wharton Marketing
Dec 22, 2011 · In social advertising, ads are targeted based on underlying social networks and their content is tailored with information that pertains to the ...
[51]
How effective is psychological targeting in advertising? What we can ...
Nov 20, 2017 · We found that matching the content of persuasive messages to individuals' psychological characteristics resulted in up to 40% more clicks and up ...
[52]
The effects of trait-based personalization in social media advertising
An experiment investigated the persuasiveness of social media ads personalized for different traits. Results show a more complex picture than previous findings ...
[53]
https://www.statista.com/topics/1538/social-media-marketing/
[54]
40+ Targeted Advertising Statistics: Data-Driven Marketing Insights ...
Mar 31, 2025 · Increase in conversion rates with personalized ads, 202%. ROI improvement with AI-driven ad personalization, 30% higher ROI. Click-through rate ...
[55]
80+ Must-Know Social Media Marketing Statistics for 2025
Feb 20, 2025 · Social media advertising spend is projected to reach $276.7 billion in 2025, with 83% of that ad spend total generated on mobile by 2030. Video ...Social Media Usage... · Facebook Statistics · Instagram Statistics
[56]
[PDF] Social Advertising Effectiveness Across Products: A Large
Oct 7, 2020 · The goal of our research is to identify the hetero- geneous effects of social advertising across products and to investigate how social ...
[57]
How to Measure Social Media Influence - Viral Nation
Aug 13, 2024 · You can gauge an influencer's impact by measuring the traffic they drive to your site using trackable URLs and UTM parameters on their posts. ...The Real Measure of Social... · How Can You Measure Social... · EngagementMissing: commercial | Show results with:commercial
[58]
10 Influencer Marketing Metrics & KPIs to Track Success - QuickFrame
Oct 2, 2024 · Key influencer marketing metrics include reach, impressions, audience engagement, follower growth, brand mentions, and conversions.
[59]
5 creator marketing metrics that matter most in 2025 - Ad Age
May 20, 2025 · The 5 creator marketing metrics that matter most in 2025—here's what to know · Reach and impressions · Engagement rate · Earned media value (EMV).
[60]
What Was Klout and Why Did It Matter? - Crazy Egg
Jan 29, 2025 · Klout was an analytics company that gave social media users a “Klout Score” between 1 and 100 based on their activity on social media platforms.
[61]
Wanna' Know How Your Klout Score is Really Calculated?
Oct 31, 2015 · Whether you agree with it or not, whether you like it or not, Klout score is used by many as a measure to determine a level of social influence.
[62]
9 Social Media Marketing Tools For Tracking KPIs and Metrics
Aug 8, 2022 · 9 Social Media Marketing Tools For Tracking KPIs and Metrics · Facebook Pixels · Google Analytics Pixel · Instagram Insights · Traackr · Hootsuite ...
[63]
Influencer Marketing Benchmark Report 2025
Apr 25, 2025 · This report delivers an unparalleled synthesis of market sentiment, investment trends, and actionable recommendations.
[64]
Everything You Need to Know About Online Reputation Score - Radarr
Discover an online reputation score by consolidating data from reviews, social media, and search visibility. It's an assessment of an entity's digital standing ...
[65]
Influencer Marketing Measurement: KPIs, Metrics, ROI - Meltwater
Jan 1, 2025 · The earned media value metric measures the value of reach and engagement received by an influencer on a campaign. It estimates how much you ...
[66]
Reputation Score: A Better Approach to Data-Driven Marketing
Holistically, Reputation Score acts as a trusted guide for business performance. It detangles complex data points from every touchpoint between your customers ...
[67]
How to Reputation Scoring
Jul 2, 2025 · Your reputation score is a numerical value based on how your brand is perceived online. You can use your reputation score to gain insight into your strengths ...
[68]
How Social Media Screenings Affect Hiring Decisions
Sep 16, 2025 · Social media screening has become a modern recruitment trend. According to a 2023 ResumeBuilder survey, 73 percent of hiring managers use social ...
[69]
Social Media Background Checks | Social Media Screening USA
Mar 22, 2025 · With 85% of recruiters now incorporating social media screening into their hiring process, businesses are increasingly relying on these checks ...
[70]
How people perceive different types of social media screening and ...
A survey revealed that 87% of recruiters identified LinkedIn as the most effective social media profile to screen job applicants during the hiring process, ...Missing: evidence | Show results with:evidence
[71]
Stop Screening Job Candidates' Social Media
According to a 2018 CareerBuilder survey, 70% of employers check out applicants' profiles as part of their screening process, and 54% have rejected applicants ...
[72]
Full article: Temporal Impacts of Problematic Social Media Content ...
Jun 18, 2023 · Job applicants' social media postings and presence can impact employers' perceptions during the hiring process.
[73]
[PDF] On the Rise of FinTechs: Credit Scoring Using Digital Footprints - FDIC
Digital footprints, from website access, can predict consumer default, and are used by FinTechs like ZestFinance and Kreditech for lending decisions.
[74]
[PDF] Credit Scoring with Social Network Data
This study analyzes how social network data impacts credit score accuracy, and how individuals may strategically modify their networks to improve scores. ...
[75]
Innovative Credit Risk Assessment: Leveraging Social Media Data ...
This study introduces a novel approach leveraging social media analytics and advanced machine learning techniques to assess the creditworthiness of individuals
[76]
(PDF) CONSUMER LENDING USING SOCIAL MEDIA DATA
The duo noted that the approach can offer an alternative in consumer credit scoring and lending process as social network data continues to grow as more and ...
[77]
Interactive Social Media: The Value for Law Enforcement | FBI - LEB
Sep 3, 2013 · Law enforcement's use of interactive social media is effective. The Philadelphia Police Department is a leader in the use of Twitter, Facebook, ...
[78]
DHS to Begin Screening Aliens' Social Media Activity for Antisemitism
Apr 9, 2025 · ... Threats, DHS will enforce all relevant immigration laws to the maximum degree, to protect the homeland from extremists and terrorist aliens ...
[79]
Homeland Security chief says social media used in immigration vetting
Dec 17, 2024 · ... threats. The Department of Homeland Security has been criticized over reports it did not routinely consult social media during the vetting ...
[80]
Full article: Predictive Policing: Review of Benefits and Drawbacks
Feb 12, 2019 · The empirical evidence provides little support for the claimed benefits of predictive policing. Whereas some empirical studies conclude that ...Research Method · Evidence For These Benefits · Claimed Drawbacks
[81]
[PDF] Social Media Use in Law Enforcement:
81% of law enforcement use social media, 25% daily, for crime prevention and investigation. 67% find it valuable for anticipating crimes, and 73% for solving ...
[82]
Why Police Should Monitor Social Media to Prevent Crime - WIRED
Apr 20, 2018 · After all, social media sites are full of data that can make police interventions more effective, from posts about crimes in progress to ...<|separator|>
[83]
Mining Social Media Data for Policing, the Ethical Way
Apr 26, 2018 · Social media posts are full of data that, when made accessible to governments, can make interventions quicker, more effective, ...
[84]
[PDF] CHINA'S SOCIAL CREDIT SYSTEM IN 2021
May 9, 2022 · Official discussions of SoCS have stressed the need to overcome data islands. Regulations and a number of Memoranda of Understanding (MoUs) ...Missing: details | Show results with:details
[85]
Social Credit Action in 2025 - China Law Translate —
Jun 11, 2024 · Last week China released a Social Credit Action Plan for 2024-2025 (the Plan). The Plan was released by the National Development and Reform ...Missing: details facts
[86]
China's Corporate Social Credit System: Context, Competition ...
Dec 8, 2020 · Officially launched in 2014 and still evolving rapidly, China's ambitious Corporate Social Credit System (CSCS) comprises a nation-wide ...
[87]
China Social Credit System Explained - How It Works [2025]
Feb 11, 2024 · Learn what the China social credit system, how it impacts individuals & companies in China and how it is operating in 2024.Timeline of the Development... · What Is the Difference... · How to Prepare for the...
[88]
China's Corporate Social Credit System and Its Implications | FSI
This research finds that politically-connected firms received higher CSCS scores, primarily by accumulating soft merits for charitable donations sanctioned by ...
[89]
China's Social Credit System: What Businesses Need to Know
Mar 24, 2023 · These measures, which will come into effect on May 1, 2023, explicitly state that companies that meet certain requirements have the right to ...Missing: facts | Show results with:facts
[90]
Impacts of China's social credit reform on firm investment efficiency
China's Social Credit Reform (CSCR) significantly enhances corporate investment efficiency. · CSCR improves efficiency via reduced agency costs, better financial ...
[91]
Does the social credit system construction reduce enterprises ...
Feb 7, 2025 · The main findings show that the social credit system construction alleviates enterprises' overinvestment behaviors, realized by the inhibiting effect of ...
[92]
Aadhaar and the Social Credit System: Personal Data Governance ...
Apr 10, 2022 · The Social Credit System in China and the Aadhaar system in India are rapid captures of personal data, and this article compares them.<|control11|><|separator|>
[93]
China's social credit system 'could interfere in other nations ...
Jun 27, 2018 · System criticised as an Orwellian tool of mass surveillance is shaping behaviour of foreign businesses, report says.
[94]
Strategic social media marketing: An empirical analysis of ...
Dec 1, 2023 · The authors study how social media can help improve sales through attention (awareness) and endorsement (engagement). Using social media ...Missing: profiling | Show results with:profiling
[95]
The value of getting personalization right—or wrong—is multiplying
Nov 12, 2021 · Research shows that personalization most often drives 10 to 15 percent revenue lift (with company-specific lift spanning 5 to 25 percent, driven ...
[96]
Credit Scoring with Social Network Data | Marketing Science
Oct 26, 2015 · Social Media Credit Scoring · Risk transfer versus cost reduction on two-sided microfinance platforms. 9 May 2018 | Quantitative Marketing and ...
[97]
Social Media Profiling in Credit Scoring - RiskSeal
Feb 25, 2025 · Social media digital credit scoring evaluates a borrower's creditworthiness by analyzing their activity on social networks. Social media ...
[98]
Layoffs and Team Restructuring: How Social Media Screening Helps
Mar 25, 2025 · Social media screening helps companies save money by reducing employee turnover. By ensuring that new hires fit well with the company culture ...
[99]
The Role of Artificial Intelligence in Personalizing Social Media ... - NIH
This paper explores the role of artificial intelligence (AI) in personalizing social media marketing strategies and its impact on customer experience.Missing: profiling | Show results with:profiling
[100]
What Consumers Want from Personalization | BCG
Dec 12, 2024 · The three benefits of personalization that respondents cited most frequently are value, enjoyment, and convenience. Two-thirds of consumers have ...
[101]
Predictive Policing: Using Technology to Reduce Crime | FBI - LEB
Apr 9, 2013 · The predictive policing model helped to alert officers to targeted locations in real time, a significant improvement over traditional tactics.
[102]
The Promises and Perils of Predictive Policing
May 22, 2025 · Research from the McKinsey Global Institute suggests that integrating AI into law enforcement might lower urban crime rates by 30 to 40 percent.
[103]
The Effect of AI-Enabled Credit Scoring on Financial Inclusion
We found that the AI model enhanced financial inclusion for the underserved population by simultaneously increasing the approval rate and reducing the default ...
[104]
(PDF) Evaluating the Role of Social Media Data in Detecting ...
Jan 27, 2025 · This study evaluates the role of social media data in detecting financial fraud patterns, exploring its potential to complement traditional detection methods.
[105]
Empirical data on the privacy paradox - Brookings Institution
The contemporary debate about the effects of new technology on individual privacy centers on the idea that privacy is an eroding value.Missing: evidence | Show results with:evidence
[106]
'Privacy Paradox' in the Social Web: The Impact of Privacy Concerns ...
This study investigates the potential influence of privacy concerns, psychological traits, attitudes to the Social Web and age on self-disclosure.
[107]
A longitudinal analysis of the privacy paradox - Sage Journals
Jun 4, 2021 · The privacy paradox states that people's concerns about online privacy are unrelated to their online sharing of personal information.<|separator|>
[108]
Discrimination in the Age of Algorithms | Journal of Legal Analysis
Apr 22, 2019 · The algorithm, unlike the human being, has no intrinsic preference for discrimination, and no ulterior motives.
[109]
Fairness perceptions of algorithmic decision-making: A systematic ...
Oct 10, 2022 · Algorithms can reduce human biases in decision-making processes because they do not grow tired, have no agency, and are not distorted by ...
[110]
Biased Humans, (Un)Biased Algorithms? | Journal of Business Ethics
Feb 28, 2022 · Previous research has shown that algorithmic decisions can reflect gender bias. The increasingly widespread utilization of algorithms in ...<|separator|>
[111]
[PDF] An exploratory study of mining social data for credit scoring
Recently, there has been tremendous interest in harnessing social media data for personal credit scoring, especially with the fast growth of a new business ...
[112]
Full article: The Effectiveness of Big Data-Driven Predictive Policing
In this study, we aimed to investigate the effectiveness of big data-driven predictive policing, one of the latest forms of technologybased policing.
[113]
Clearview AI fined $33.7 million by Dutch data protection watchdog ...
Sep 3, 2024 · The Dutch data protection watchdog on Tuesday issued facial recognition startup Clearview AI with a fine of 30.5 million euros ($33.7 million)
[114]
The French SA fines Clearview AI EUR 20 million
On 26 November 2021, the Chair of the CNIL decided to give Clearview AI formal notice to cease the collection and use of data of persons on French territory in ...
[115]
Data Brokers – EPIC – Electronic Privacy Information Center
Data brokers also threaten national security by compiling and selling extensive profiles of information on members of the military and government officials.
[116]
[PDF] DATA BROKERS AND SECURITY
The tech-journal Comparitech recently reported on a data brokerage that accidentally exposed its database of 235 million social media profiles, which included ...
[117]
Data Broker Breaches - Insider Threats and More - Hornetsecurity
Sep 19, 2024 · The massive data breach at data broker National Public Data exposed over 2.9 billion personal information records. They discuss the risks of ...
[118]
How do data brokers affect the threat ecosystem? - Barracuda Blog
May 2, 2025 · When a data breach occurs, hackers can use data broker databases to enrich stolen data with more information. For example, breaching a Zoom ...
[119]
Federal Government Social Media Surveillance, Explained
Feb 9, 2022 · ... invasions of privacy, and sometimes high-stakes misinterpretations of social media activity. While all Americans may be harmed by ...
[120]
China's Social Credit System in 2021: From fragmentation towards ...
Mar 3, 2021 · On the technical side, the collection of increasing amounts of data in the context of the SoCS means higher risks of data leaks and abuse of ...
[121]
Closing the Data Broker Loophole | Brennan Center for Justice
Feb 13, 2024 · This underregulated data broker ecosystem, built to target consumers with ads, also engenders ever-increasing risks of data breaches; ...
[122]
Algorithmic bias detection and mitigation: Best practices and policies ...
May 22, 2019 · Our research presents a framework for algorithmic hygiene, which identifies some specific causes of biases and employs best practices to ...
[123]
Machine Bias - ProPublica
May 23, 2016 · There's software used across the country to predict future criminals. And it's biased against blacks.Missing: Kleinberg | Show results with:Kleinberg
[124]
[PDF] LESSONS FROM THE COMPAS- PROPUBLICA DEBATE
ProPublica investigative journalists claimed that the COMPAS algorithm is biased and released their findings as open data sets. The ProPublica data started a ...
[125]
Bias in Criminal Risk Scores Is Mathematically Inevitable ...
Dec 30, 2016 · An article published earlier this year by ProPublica focused attention on possible racial biases in the COMPAS algorithm. We collected the ...
[126]
[PDF] Algorithmic Bias in Law Enforcement's Facial Recognition ...
Feb 8, 2021 · TECH. & INNOVATION. FOUND., THE CRITICS WERE WRONG: NIST DATA SHOWS THE BEST FACIAL. RECOGNITION ALGORITHMS ARE NEITHER RACIST NOR SEXIST 2 ( ...
[127]
Algorithmic Bias? An Empirical Study into Apparent Gender-Based ...
This study examines some potential explanations for gender discriminatory outcomes produced from advertising algorithms.
[128]
No nonsense version of the "racial algorithm bias" - LessWrong
Jul 13, 2019 · Northpointe showed that COMPAS is approximately fair in calibration for Whites and Blacks. ProPublica showed that COMPAS is unfair in parity.
[129]
Questions and answers on the Digital Services Act*
The new rules only impose measures to remove or encourage removal of illegal content, in full respect of the freedom of expression. At the same time, the DSA ...
[130]
Is the EU's Digital Services Act Compliant with The Right to Freedom ...
Feb 14, 2025 · The DSA's application is likely to result in limitations to free speech, amounting to a potential infringement of Article 11 of the EU Charter and Article 10 ...<|separator|>
[131]
The EU Digital Services Act Could Cripple Free Speech – Even In ...
Aug 29, 2025 · It may sound like an innocuous, boring European law, but the DSA is a digital gag order with global consequences. If left unchecked, it can ...Missing: criticism | Show results with:criticism
[132]
https://itif.org/publications/2025/10/20/eu-digital-services-act-lacks-transparency-to-assess-its-impact-on-free-speech-globally/
[133]
Digital Services Act: Does Internet Regulation Threaten Freedom of ...
Jul 21, 2025 · Johanna Rinceanu critiques the Digital Services Act, highlighting risks to freedom of expression posed by tech companies' unchecked power.
[134]
Why the Government Should Not Regulate Content Moderation of ...
Apr 9, 2019 · Regulations of social media companies might either indirectly restrict individual speech or directly limit a right to curate an internet ...
[135]
Social media court case shows we must rein in bureaucrats to ...
Jul 3, 2024 · Missouri disappointed many Americans concerned about federal bureaucrats secretly coercing social media companies to censor free speech online.
[136]
Summarizing the Section 230 Debate: Pro-Content Moderation vs ...
Jul 5, 2022 · The debate surrounding online content moderation, which is governed by Section 230 of the Communications Decency Act of 1996.
[137]
FIRE Report on Social Media 2024
As the First Amendment's prohibition on censorship makes clear, the government itself poses the greatest threat to free speech rights. This is no different with ...<|separator|>
[138]
The Potential Impact of Proposed Changes to Section 230 on ...
Section 230, a law that establishes critical liability protection for a range of online services that carry users' content, has faced an increasing amount of ...Missing: overreach | Show results with:overreach
[139]
Leveraging Artificial Intelligence and Machine Learning to Profile ...
Feb 1, 2024 · Artificial Intelligence and Machine Learning have progressed enough to be able to profile internet users through data collected via social media ...
[140]
Multimodal Generative AI and Foundation Models for Behavioural ...
This narrative review examines how artificial intelligence, particularly multimodal generative models and foundation technologies, can address these issues by ...
[141]
Web-Browsing LLMs Can Access Social Media Profiles and Infer ...
Jul 16, 2025 · We show that these models can access social media content and predict user demographics with reasonable accuracy.
[142]
High-level summary of the AI Act | EU Artificial Intelligence Act
AI systems listed under Annex III are always considered high-risk if it profiles individuals, i.e. automated processing of personal data to assess various ...Prohibited Ai Systems... · High Risk Ai Systems... · General Purpose Ai (gpai)Missing: accountability | Show results with:accountability
[143]
AI Act: Prohibited AI Practices Become Applicable | EY - Greece
Feb 4, 2025 · Under Article 5(1)(c) of the AI Act, the prohibition of social scoring is broadly applicable to both public and private entities that implement ...
[144]
Artificial intelligence (AI) act: Council gives final green light to the ...
May 21, 2024 · With the AI act, Europe emphasizes the importance of trust, transparency and accountability when dealing with new technologies while at the ...
[145]
Risk-Based AI Regulation: A Primer on the Artificial ... - RAND
Nov 20, 2024 · The EU AI Act bans certain AI systems deemed to pose unacceptable risks, imposes extensive requirements on high-risk systems, and defines ...Risk-Based Ai Regulation · Risk Framework For The Eu Ai... · Who Enforces The Eu Ai Act?
[146]
[PDF] Ethics, Transparency and Accountability Framework for Automated ...
May 13, 2021 · What the framework is for. Understanding automated and algorithmic decision-making. Before you use this framework. How to use this framework.
[147]
Predictive Policing Explained | Brennan Center for Justice
Apr 1, 2020 · Predictive policing involves using algorithms to analyze massive amounts of information in order to predict and help prevent potential future crimes.
[148]
Predictive Policing Examples and Case Studies - IGNESA
Aug 5, 2024 · Implementing accountability mechanisms, such as oversight boards and external reviews, to monitor the use of predictive policing technologies.
[149]
Exploring the tradeoff between data privacy and utility with a clinical ...
May 30, 2024 · This study aimed to demonstrate the effect of different de-identification methods on a dataset's utility with a clinical analytic use case
[150]
Surveillance and Predictive Policing Through AI - Deloitte
Cities already capture images for surveillance purposes, but by using AI images can now be analysed and acted on much more quickly. Machine learning and big ...
[151]
Social Accountability: What Does the Evidence Really Say?
Aug 5, 2025 · Empirical evidence of tangible impacts of social accountability initiatives is mixed. This meta-analysis reinterprets evaluations through a ...