Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces the concept of "functional misalignment" to explain adverse societal outcomes arising from AI-driven digital platforms, despite their success in predicting individual behavior. It posits that optimizing for easily observable signals like clicks leads to a bias against reflective judgment, creating feedback loops and emergent collective dynamics that amplify negative effects. The paper then proposes a research agenda for studying and mitigating these misalignments.
Social media's troubling side effects, like polarization and declining mental health, aren't glitches but features arising from algorithms that are *too good* at predicting (and exploiting) our behavior.
Algorithmic systems, particularly social media recommenders, have achieved remarkable success in predicting behavior. By optimizing for observable signals such as clicks, views, and engagement, these systems effectively capture user attention and guide interaction. Yet their widespread adoption has coincided with troubling outcomes, including rising mental health concerns, increasing polarization, and erosion of trust. This paper argues that these effects are consequences of a structural functional misalignment between what algorithms optimize - predictable behavior - and the human goals these predictions are intended to serve. We propose that this misalignment arises through three mechanisms: (1) a bias toward modeling fast, reactive behavioral signals over reflective judgment, (2) feedback loops that couple user behavior with algorithmic learning, and (3) emergent collective dynamics that amplify these effects at scale. Together, these mechanisms explain how accurate individual-level predictions can produce adverse societal outcomes. We present functional misalignment as a unifying framework and outline a research agenda for studying and mitigating its effects in human-AI interaction systems.