Your phone doesn’t know you intuitively. It knows you statistically. That distinction matters more than most people realize.
Sandra Matz’s Mindmasters: The Data-Driven Science of Predicting and Changing Human Behavior makes that case with precision. Matz, a professor at Columbia Business School, has spent years studying psychological targeting: the practice of using behavioral data and AI to predict what people want, fear, believe, and will do next. The book is part scientific briefing, part ethical argument, and part user manual for a world where your digital behavior is being read more carefully than you read yourself.
This review covers what psychological targeting actually is, how AI builds profiles from behavioral data, the ethical questions Matz raises about data ownership, and what practical steps she recommends.
What You’ll Learn
- How AI systems build psychological profiles from behavioral data
- Why psychological targeting is both a tool for empowerment and a mechanism for manipulation
- What data cooperatives are and why Matz believes they matter
- How to recognize if you’re being profiled
- Practical steps to reduce your exploitable data footprint
What Is Psychological Targeting and How Does AI Use It?
Psychological targeting is the practice of using behavioral and demographic data to infer personality traits, values, and vulnerabilities, then crafting messages designed to influence specific individuals. AI systems apply this at scale by analyzing patterns across millions of data points.
The underlying research is real. A 2013 study from Cambridge University found that Facebook likes alone could predict personality traits with statistically significant accuracy. Patterns in what people engage with—not just what they say—reveal introversion, political leanings, and emotional tendencies. Matz builds on this research to explain how the same methods now operate inside advertising systems, political campaigns, financial institutions, and hiring platforms.
The implication is worth sitting with. When you interact with a platform, you are not just consuming content. You are generating a behavioral record that feeds models designed to predict and shape what you do next.
Key takeaway: Psychological targeting isn’t a fringe practice. It’s embedded in the systems most people use daily, and its accuracy improves as behavioral data accumulates.
How Does AI Build a Psychological Profile From Your Behavior?
AI systems construct psychological profiles by aggregating behavioral signals across platforms and time—clicks, searches, engagement patterns, purchase history, and social connections—then mapping those signals onto psychological models that predict personality and decision-making tendencies.
The process doesn’t require you to disclose anything directly. Behavior is the disclosure. Every search query, every content interaction, every pattern of engagement contributes data points. The algorithms infer from correlation, not confession.
Matz is specific about what these systems can derive: personality traits across the OCEAN model (openness, conscientiousness, extraversion, agreeableness, neuroticism), political inclinations, mental health indicators, and financial vulnerability. This level of inference creates profiles that are often more detailed than what someone would share with a close friend.
The concerning part isn’t just the accuracy. It’s the asymmetry. The organizations building these profiles understand the implications; most people whose data feeds them do not.
Key takeaway: AI psychological profiling doesn’t require self-disclosure. Behavioral patterns are sufficient, and the resulting profiles can be more accurate than the person’s own self-assessment.
Is Psychological Targeting Always Manipulative?
Psychological targeting is a tool. Whether it functions as manipulation or genuine service depends on who controls it, toward what end, and with what transparency.
Matz resists the purely dystopian framing, and the book is stronger for it. She presents research from her own lab showing that AI-matched financial messaging, tailored to personality type, helped people save more money. Impulsive spenders responded to messages framing savings as immediate wins. Long-term planners responded to messages about future security. The same underlying intervention, calibrated differently, produced meaningfully better outcomes.
The Cambridge Analytica case sits on the opposite end of that spectrum. There, the same targeting infrastructure was used to exploit psychological vulnerabilities in service of political manipulation, without the knowledge or consent of the people being targeted. Matz uses it not as a headline but as a structural illustration: the technology enables both outcomes, and the difference between them is intent, consent, and accountability.
The ethical line she draws is clearer than most treatments of this topic. Targeting becomes manipulation when it bypasses understanding in order to produce behavior. It functions as a service when it improves outcomes for the person being targeted, with their awareness.
Key takeaway: Psychological targeting becomes manipulation when it exploits vulnerabilities without consent. The tool itself is neutral; the ethical question is whether it serves or bypasses the person it’s applied to.
Who Controls Your Behavioral Data, and Should That Change?
Right now, the behavioral data that feeds psychological targeting systems is controlled almost entirely by the platforms that collect it. Users generate the data and bear the consequences of its use, but have limited visibility into how it’s applied and minimal legal recourse in most jurisdictions.
Matz’s proposed solution is data cooperatives: collectively owned data structures that function like digital credit unions, where members decide collectively how their data is used and who benefits from it. The idea is structurally sound. Instead of each individual negotiating alone against organizations with massive information and legal advantages, a cooperative would pool both data and decision-making power.
The honest critique is that this remains more architectural proposal than operational reality. The regulatory infrastructure to mandate or support data cooperatives doesn’t yet exist at scale. And the corporations that would need to cede control over that data have significant incentives to resist it.
What Matz accomplishes, even if the cooperative model stays aspirational, is reframe the question. Data ownership is not a technical problem. It’s a governance problem. And governance problems are solvable when enough people treat them as such.
Key takeaway: Data cooperatives represent a structural response to the asymmetry between individuals and platforms. Whether they become viable depends on regulatory will and collective organization, not just technology.
How Can You Tell If You’re Being Psychologically Profiled?
You almost certainly are. The more useful question is where and to what effect.
The signals Matz describes are recognizable: advertising that feels specific to concerns you haven’t articulated, content recommendations that consistently reinforce existing beliefs, financial products or job offers that seem calibrated to personal characteristics rather than declared preferences.
More troubling are the cases she raises where profiling extends into consequential decisions. Insurance companies exploring psychological data to assess risk. Banks testing personality-based assessments for loan approval decisions. These applications move profiling out of the attention economy and into access to financial and social infrastructure.
Rule of thumb: If an offer or recommendation feels more personally attuned than it should based on what you’ve explicitly shared, behavioral inference is likely involved.
Common mistake: Assuming that using privacy settings resolves the issue. Behavioral data collection often persists across platforms and devices even when ad personalization is turned off.
What Can You Actually Do About Psychological Profiling?
Matz is careful not to offer a false sense of complete protection, but she identifies meaningful leverage points. Reducing your exploitable data footprint doesn’t require disappearing from the internet. It requires reducing the signal quality available to systems designed to model you.
Her practical recommendations:
- Turn off ad personalization across major platforms. This reduces the signal feedback loop, even if it doesn’t eliminate profiling entirely.
- Use privacy-focused tools. Search engines like DuckDuckGo and tracker-blocking browser extensions reduce passive behavioral data collection.
- Exercise existing data rights. GDPR in Europe and the California Consumer Privacy Act in the US give individuals meaningful, if limited, rights to access, correct, and delete personal data. Most people never use them.
- Support regulatory pressure for AI transparency. The structural asymmetry Matz describes doesn’t resolve through individual behavior alone. Organizations that push for algorithmic accountability are working on the governance layer where real change happens.
None of this makes you invisible. But the goal isn’t invisibility. The goal is reducing the depth of the profile that can be built without your knowledge or consent.
Key takeaway: Meaningful privacy protection combines individual behavior change with support for the regulatory and governance structures that address the systemic problem. Neither alone is sufficient.
What Does Mindmasters Get Right—and Where Does It Fall Short?
Mindmasters succeeds at its primary task: making the research accessible and the stakes legible without oversimplifying either.
Matz writes for a general audience without condescending to one. The empirical grounding is strong. She draws on her own published research as well as a wide body of work in behavioral science, and she’s careful to distinguish between what the data shows and what she believes it implies. That distinction—observation versus recommendation—makes the book more credible and more useful.
The weaker sections involve the proposed solutions. Data cooperatives are described as a promising structural alternative, but the path from here to there remains underspecified. Readers looking for policy prescriptions will find the analysis sharper than the remedies.
The book also doesn’t fully engage with the speed of change in this space. AI capabilities in psychological inference have continued to advance since Matz was writing, and some of the examples that felt like outliers at time of publication are closer to standard practice now.
Still, Mindmasters is a better-than-average entry in a crowded category of AI ethics books. It’s less alarming and more analytical than most. That balance makes it more useful.
Is Mindmasters Worth Reading?
Mindmasters is worth reading for two reasons that have nothing to do with novelty. First, Matz explains the actual mechanism behind psychological targeting clearly enough that readers come away with genuine understanding, not just ambient concern. Second, her framing of the governance problem is more useful than most popular treatments.
If you work in marketing, advertising, or product design, this book is closer to required reading. The systems Matz describes aren’t external to those fields. They are those fields, increasingly. Understanding what they actually do—and what distinguishes ethical use from exploitation—is a professional responsibility, not just a personal one.
If you’re a general reader trying to understand why the digital environment feels designed to move you in directions you didn’t choose, Mindmasters gives you the language and the framework to think about it clearly.
