How to “improve” prediction using behavior modification

Many internet platforms that collect behavioral big data use it to predict user behavior for internal purposes and their business customers (e.g., advertisers, insurers, security forces, governments, political consulting firms) who utilize the predictions personalization, targeting, other decision-making. Improving predictive accuracy is therefore extremely valuable. Data science researchers design algorithms, models, approaches improve prediction. Prediction also improved with larger richer data. Beyond improving algorithms data, can stealthily achieve better prediction by pushing users’ behaviors towards predicted values, using modification techniques, thereby demonstrating more certain predictions. Such apparent “improved” result from employing reinforcement learning combine modification. This strategy absent machine statistics literature. Investigating its properties requires integrating causal notation. To this end, we incorporate Pearl’s do(.) operator into vocabulary. We then decompose expected error given identify components impacting power. Our derivation elucidates implications of such scientists, platforms, customers, humans whose manipulated. Behavior make predictable even homogeneous; yet predictability might not generalize when in practice. Outcomes pushed be at odds customers’ intentions, harmful manipulated users.