In 2013, researchers at the University of Cambridge published a landmark study analysing 58,000 Facebook users. They discovered that "liking" content predicted Big Five personality scores with meaningful accuracy. Extraversion, Openness, and Neuroticism showed the strongest correlations, with Extraversion reaching approximately .40. The study, by Kosinski, Stillwell, and Graepel (doi:10.1073/pnas.1218772110), demonstrated that passive digital behaviour carries real personality signal — and opened a set of questions about what that means for inference, privacy, and the ethics of personality data.
What Big Five Research Shows About Personality Signals in Social Media
Extraversion (Presence). High-extraversion users post more frequently, maintain larger social networks, and engage more with others' content. Effect sizes typically range from r = .30 to .45, making Presence the most reliably detectable Big Five dimension in social media data. For a full account of what Presence involves, see what Extraversion means beyond the introvert-extrovert binary.
Openness to Experience (Vision). High-openness individuals engage with diverse content, share articles about cultural or intellectual topics, and use more complex vocabulary. Text-based predictions yield correlations of approximately .30 to .40. For more on how Vision manifests in creative and intellectual behaviour, see creativity and personality: what Big Five research shows.
Neuroticism (Depth). Moderately predictable. High-neuroticism users post more during evenings and late nights, use negative emotional vocabulary, and show variable posting frequency (r = .20–.35). For a deeper understanding of Depth as a dimension, see what Neuroticism means at work.
Conscientiousness (Discipline). Harder to detect. High-conscientiousness users maintain complete profiles and post regularly, but signal strength is weaker (r = .15–.25).
Agreeableness (Bond). The hardest trait to detect. High-agreeableness users behave cooperatively online, but signals are subtle and difficult to distinguish from other traits (r = .10–.20).
The Cambridge Analytica Case: How Personality Research Became Manipulation
Researcher Aleksandr Kogan built a Facebook quiz app collecting personality data from users and their friends without explicit consent. Cambridge Analytica obtained this data and claimed to use psychographic profiling for political targeting in the 2016 US election and Brexit referendum.
This case demonstrated two critical points: the gap between academic research and commercial manipulation was smaller than anticipated, and consent frameworks for behavioural data needed fundamental rethinking. The Cambridge Analytica scandal became a defining moment in public awareness of how personality data can be weaponised — and how the correlations that are modest at the individual level can still be exploited at population scale.
However, the academic correlations are modest enough that individual-level targeting based on personality profiles introduces substantial noise. The scandal amplified legitimate concerns about consent and data use that the research alone may not have warranted. For a related discussion of how personality assessments can be gamed or misused, see can you fake a personality test and social desirability bias in personality tests.
What Social Media Posts Cannot Reliably Reveal About Personality
Social media posts reliably predict aggregate personality tendencies but fail to predict:
Specific intentions and decisions. Personality traits explain long-run tendencies, not specific acts. High-extraversion scores don't predict whether someone will attend a particular event or vote a certain way. For a detailed treatment of what personality science cannot conclude about individuals, see personality science: limits and what it cannot predict.
Context-sensitive behaviour. People express personality differently across contexts. Professional LinkedIn activity may differ substantially from personal Twitter accounts, not because traits changed but because they manifest differently contextually.
Individual accuracy. Correlations like r = .35 or r = .40 translate into substantially weaker individual-level prediction accuracy than these numbers suggest. Uncertainty around personality inferences from social media remains large. The replication and accuracy concerns in personality science more broadly are addressed in the personality science replication crisis.
| Social media signal | Big Five dimension predicted | Typical correlation estimate |
|---|---|---|
| Posting frequency, social content | Extraversion (Presence) | r ≈ .35–.45 |
| Content diversity, cultural engagement | Openness (Vision) | r ≈ .30–.40 |
| Late-night posting, emotional vocabulary | Neuroticism (Depth) | r ≈ .20–.35 |
| Profile completeness, posting regularity | Conscientiousness (Discipline) | r ≈ .15–.25 |
| Cooperative tone, conflict avoidance | Agreeableness (Bond) | r ≈ .10–.20 |
Why Cèrcol Uses Witness Ratings Instead of Social Media Signals
The research demonstrates that behavioural data carries personality information but does not demonstrate that such inference is accurate, fair, or ethically grounded for individual assessment. Cèrcol's model relies on explicit participation: self-report combined with peer assessment from Witnesses who know the individual in context.
The distance from "detectable in aggregate social media data" to "appropriate for individual assessment" is larger than typically acknowledged. Social media inference bypasses the individual's knowledge and consent; it draws on context-collapsed, platform-shaped behaviour that may not reflect how a person actually shows up in the contexts that matter for their work. The Witness peer assessment takes the opposite approach: structured observations from people who have worked directly with the individual, in the specific contexts where personality matters most.
For a comparison of assessment approaches and their validity, see Cèrcol's science page, which explains the psychometric foundations underlying both instruments.
Know Your Own Profile — On Your Own Terms — with Cèrcol
The insight from social media personality research is that your digital behaviour reflects your personality whether you intend it to or not. But there is a significant difference between having your personality inferred from your posts without your knowledge, and actively choosing to understand your own profile through a validated assessment that you control.
Cèrcol's free Big Five assessment gives you a precise profile across all five dimensions — Presence, Vision, Depth, Discipline, and Bond — based on 120 validated items and your own deliberate participation. You can see exactly what the data shows and use it on your own terms. The Witness peer assessment adds a layer of behavioural observation from colleagues who know you in context — far more informative than any passive social media inference.
Take the free Cèrcol assessment at cercol.team
Sources
- Kosinski, M., Stillwell, D., & Graepel, T. (2013). Private traits and attributes are predictable from digital records of human behavior. PNAS, 110(15), 5802–5805. doi:10.1073/pnas.1218772110
- Facebook–Cambridge Analytica data scandal — Wikipedia
- Big Five personality traits — Wikipedia
Further reading
- What is Extraversion? Beyond the introvert-extrovert binary
- Creativity and personality: what Big Five research shows
- What is Neuroticism? Understanding emotional depth at work
- Can you fake a personality test?
- Social desirability bias in personality tests
- Personality science: limits and what it cannot predict
- Personality science replication crisis