Towards a New Psychology of Human-AI Interaction (javier-marin.medium.com)

The piece argues that the classic psychological and HCI paradigms—born from Tversky and Kahneman’s insights on human bias and the 1980s’ user-centered design work of Card, Moran, Newell, Norman and Shneiderman—are reaching an inflection point as AI systems stop acting merely as tools and become autonomous, adaptive partners. Where earlier models treated computers as predictable instruments for human intent, contemporary ML-driven agents shape attention, choices and habits through personalization and feedback loops. The author frames this as a projective claim: the technologies we build will reshape cognition and identity over time, echoing Heraclitus’ notion that our repeated choices become who we are. For AI/ML practitioners and researchers this shift matters because it demands new theory, measurement and design methods. Technically, we must move beyond static usability and accuracy metrics to models that capture bidirectional influence: how algorithms learn from behavior while simultaneously altering it (path dependence, preference drift, reinforcement of biases). That implies longitudinal experiments, causal inference tools to separate algorithmic effects from organic change, and interdisciplinary frameworks combining ML, behavioral science and ethics. Practically, designers, evaluators and regulators should prioritize metrics like autonomy, resilience to manipulation and long-term cognitive impact, not just short-term engagement or performance.