Psychographic Profiling in the Age of AI

Groundbreaking research by Michal Kosinski demonstrated that digital footprints can predict personality traits with remarkable precision. For instance, analyzing as few as ten Facebook likes enables predictions about personality traits more accurately than those of a coworker. With seventy likes, the model surpasses the understanding of close friends. At one hundred and fifty likes, it outperforms family members, and with three hundred likes, it rivals a spouse’s knowledge (Kosinski et al., 5803). These findings underscore how digital behavior analysis can outpace even the closest human relationships in understanding personality.

The ability to decode personality traits has transformative implications for persuasion. Studies leveraging the "Big Five" psychological model—Openness, Conscientiousness, Extraversion, Agreeableness, and Neuroticism (OCEAN)—show that tailoring messages to align with individual psychological profiles significantly boosts engagement. For example, individuals high in Openness respond more positively to ads emphasizing creativity and novelty, whereas those high in Neuroticism are influenced by messages focusing on security and stability. Psychologically tailored campaigns have demonstrated a 40% increase in clicks and a 50% rise in purchases compared to non-tailored approaches (Matz et al., 12715).

However, psychographic profiling also presents ethical risks, as highlighted by the Cambridge Analytica scandal in 2018. The company harvested data from millions of Facebook users, often without their consent, through a personality quiz app. This data was then used to create psychographic profiles based on the Big Five traits, enabling targeted political campaigns designed to exploit individual fears and motivations. These tactics reportedly influenced pivotal events, such as the 2016 U.S. presidential election and the Brexit referendum. This scandal emphasized the darker side of psychological profiling, including privacy violations and manipulation, which sparked global calls for stricter data protection laws and accountability for tech companies (Cadwalladr, 2018).

Scaling Persuasion with AI

Kosinski’s research laid the groundwork for understanding how digital footprints predict and influence behavior. However, the rise of large language models (LLMs), such as GPT-4, has pushed these capabilities to new heights.

Murphy (2024) highlights GPT-4’s remarkable accuracy in predicting personality traits based on digital footprints. For instance, analyzing an individual’s 50 most recent tweets allows GPT-4 to classify Myers-Briggs personality types with 76% accuracy, significantly outperforming traditional machine learning models like recurrent neural networks, which achieved 49.75% accuracy. This advancement underscores the potential of LLMs to refine psychological profiling with greater precision and efficiency (Murphy, 2024).

Goldstein et al. (2023) describe how LLMs have enabled unprecedented scalability in influence operations. For example, these models can generate dynamic, culturally appropriate, and linguistically nuanced content, overcoming limitations of earlier disinformation campaigns that often relied on repetitive, copy-pasted messaging. This capability makes AI-generated propaganda less detectable and more effective at swaying target audiences (Goldstein et al., 2023).

Critical Risks of AI-Driven Influence

1. Exploitation of Emotional Vulnerabilities

Murphy (2024) demonstrates that AI systems like GPT-4 excel at predicting nuanced personality traits, such as openness and neuroticism, enabling them to craft deeply personalized and emotionally resonant messages. For instance, AI-generated content can evoke specific emotions, such as trust, urgency, or fear, to influence behavior. Studies indicate that AI-driven models consistently outperform traditional human-created messaging, achieving higher engagement and relatability (Schmidt et al., 2022; Peters and Matz, 2023).

2. Amplification of Biases and Polarization

Social media algorithms already create filter bubbles that limit exposure to diverse perspectives (Pariser, 2011). Murphy’s research raises additional concerns about how AI-driven profiling deepens these divides by delivering emotionally charged content tailored to specific personality types. This amplification of biases contributes to ideological polarization and the spread of misinformation (Vosoughi et al., 2018; Murphy, 2024).

3. Blurred Lines Between Human and Machine Influence

Murphy (2024) highlights how the human-like sophistication of AI-generated content makes it difficult for users to distinguish between machine-crafted messaging and human communication. This poses significant ethical risks, as individuals unknowingly engage with persuasive AI content. Studies published in Nature Machine Intelligence show that users often rate AI-generated text as more relatable and convincing than human-authored messages (Schmidt et al., 2022).

4. Unprecedented Scalability and Automation

The scalability of AI-driven profiling is a key concern. Murphy (2024) notes that LLMs can analyze massive amounts of data in real time, enabling the automated delivery of micro-targeted messages to entire populations. For example, LLMs can replicate the methods used during the Cambridge Analytica scandal, but on a much larger and more precise scale, amplifying their potential impact on elections and public opinion (Cadwalladr, 2018).

Geopolitical Implications: The TikTok Debate

The potential misuse of psychological profiling extends beyond commercial and political campaigns to matters of national security. TikTok, the immensely popular social media platform owned by the Chinese company ByteDance, has come under scrutiny for its data collection practices. Critics argue that TikTok’s access to user data—including viewing habits, behavioral patterns, and preferences—could enable foreign entities to build psychological profiles of millions. These profiles could then be weaponized to spread propaganda, sway public opinion, or influence geopolitical decisions (Bennett, 2023).

Goldstein et al. (2023) further elaborate on how platforms like TikTok could exploit user data to deploy dynamic influence campaigns tailored to specific audiences. For example, AI-powered models can analyze engagement metrics to refine disinformation narratives in real-time, making them more effective and harder to detect (Goldstein et al., 2023). These concerns echo the ethical issues raised during the Cambridge Analytica scandal, showcasing how psychological profiling can have global ramifications when wielded by powerful entities.

While TikTok’s entertaining content often masks its data collection practices, the underlying concerns align with those of other platforms. These include how data can be used to manipulate behavior and exert influence on a global scale, highlighting the urgent need for greater transparency and regulatory oversight.

The Risks of Centralized Control: Musk, Oligarchs, and Democracy

The dangers of profiling and persuasion are further amplified when centralized control of digital platforms comes into play. Elon Musk’s acquisition of Twitter, now rebranded as X, illustrates the risks of concentrating power in a single individual’s hands. By owning a platform with access to real-time data on user behavior—likes, tweets, and retweets—Musk can observe and influence the emotional and psychological pulse of millions.

Critics argue that this level of control grants Musk unparalleled power to shape public discourse, manipulate markets, and even influence geopolitical debates. For example, Musk has previously used his influence on Twitter to sway cryptocurrency markets, highlighting how psychological insights combined with platform control can shape behavior at scale. More recently, Musk played a pivotal role in the 2024 election of Donald Trump by amplifying specific narratives on X. While Musk frames his ownership as a move to protect free speech, concerns remain about his ability to steer narratives through algorithmic design or direct intervention, raising critical questions about the balance of power in a digitally connected world.

These concerns echo President Joe Biden’s warning about the growing influence of oligarchs. In his farewell address, Biden cautioned against the concentration of power in the hands of a few wealthy individuals, describing the “tech-industrial complex” as a force capable of spreading misinformation and manipulating public opinion. Although Musk was not explicitly named, many interpreted Biden’s remarks as a critique of tech magnates like him, highlighting the risks posed when technological influence is centralized.

Balancing Benefits and Risks: What Can Be Done?

The use of digital profiling is not inherently harmful. Personalized experiences, improved healthcare, and better product recommendations are among its positive applications. However, without transparency and regulation, the same tools can exploit vulnerabilities, manipulate behaviors, and erode trust in democratic institutions.

Opting Out of Behavioral Targeting

Limiting the amount of behavioral data shared with platforms can significantly reduce the effectiveness of psychological profiling. Research published in the Journal of Consumer Research found that opting out of personalized advertising decreases exposure to tailored persuasive messaging, disrupting the feedback loop that enables precise targeting (Smith et al., 127).

Advocating for Algorithmic Transparency

Transparency in how algorithms collect and use data is critical for mitigating the risks of profiling. Studies argue that when users and regulators gain insight into algorithmic processes, the potential for manipulation diminishes (Zarsky, 55). Transparent systems empower individuals to make informed decisions and establish oversight to prevent exploitation.

Supporting Data Minimization Legislation

Regulations that limit the collection of personal data are crucial in curbing psychological profiling. Europe’s General Data Protection Regulation (GDPR) has proven effective in restricting unnecessary data collection, with reports showing a 20% reduction in data harvesting post-GDPR implementation (Nature Human Behaviour, 5). Advocating for similar policies globally ensures platforms gather only essential data, reducing the risk of manipulation.

Increasing Awareness through Education

Educating users about the methods and risks of psychological profiling enables individuals to recognize and resist manipulative techniques. A study in the Journal of Information Literacy revealed that digital literacy programs improved users’ ability to critically assess targeted content (Jones and Taylor, 89). By understanding how profiling works, individuals are better equipped to protect themselves from undue influence.

The Takeaway

Your online behavior is more than just a series of clicks and shares—it’s a window into your mind. From academic researchers to corporations and governments, powerful entities are leveraging this data to influence opinions and behaviors on an unprecedented scale. The rise of AI has accelerated these capabilities, making it critical to understand and manage your digital footprint. By staying informed and taking proactive steps, individuals can navigate this digital age with greater autonomy and awareness.

Bibliography

Bennett, Cy. “TikTok Ban: A National Security Concern or Political Posturing?” Digital Privacy Quarterly, vol. 12, no. 3, 2023, pp. 45-49.

Cadwalladr, Carole. “The Great Hack: How Cambridge Analytica Weaponized Data.” The Guardian, 18 Mar. 2018, www.theguardian.com/news/2018/mar/18/cambridge-analytica-facebook-influence-us-election.

Goldstein, J., McCarrick, J., & Taylor, P. (2023). Generative AI and Influence Operations: New Risks in a Digital Age. Journal of Emerging Technologies, 15(3), 112–130.

Kapoor, Kritika, et al. “Emotional Targeting in Digital Advertising: Leveraging Personality Traits to Enhance Engagement.” Journal of Digital Marketing Strategies, vol. 34, no. 2, 2021, pp. 112-127.

Kosinski, Michal, et al. “Private Traits and Attributes Are Predictable from Digital Records of Human Behavior.” Proceedings of the National Academy of Sciences, vol. 110, no. 15, 2013, pp. 5802–5805. doi:10.1073/pnas.1218772110.

Kosinski, M. (2024). Evaluating large language models in theory of mind tasks. Proceedings of the National Academy of Sciences, 121(45), e2405460121.

Matz, Sandra C., et al. “Psychological Targeting as an Effective Approach to Digital Mass Persuasion.” Proceedings of the National Academy of Sciences, vol. 114, no. 48, 2017, pp. 12714–12719. doi:10.1073/pnas.1710966114.

Obadimu, Abiodun, et al. “Using Social Media Data for Psychological Profiling: Opportunities and Challenges.” Journal of Social Media Studies, vol. 9, no. 2, 2023, pp. 67–89.

Pariser, Eli. The Filter Bubble: What the Internet Is Hiding from You. Penguin, 2011.

Peters, Brendan, and Sandra C. Matz. “Leveraging Language Models for Psychological Profiling: An Examination of Big Five Trait Inference.” Journal of Computational Psychology, vol. 15, no. 1, 2023, pp. 1-15.

Schmidt, Anne, et al. “Trust in AI-Generated Text: Implications for Persuasion and Behavioral Influence.” Nature Machine Intelligence, vol. 4, no. 8, 2022, pp. 673–681.

Smith, Andrew, et al. “Reducing Exposure to Personalized Persuasion.” Journal of Consumer Research, vol. 48, no. 3, 2021, pp. 123–129.

“The Impact of GDPR on Data Collection Practices.” Nature Human Behaviour, vol. 5, 2021, pp. 45–56.

Vosoughi, Soroush, et al. “The Spread of True and False News Online.” Science, vol. 359, no. 6380, 2018, pp. 1146–1151. doi:10.1126/science.aap9559.

Zarsky, Tal Z. “The Trouble with Algorithmic Transparency.” Harvard Journal of Law & Technology, vol. 29, no. 1, 2016, pp. 55–72.