eLetter published in Science (Vol. 391, No. 6792) on April 1, 2026, in response to Perry, A. (2026). In defense of social friction. Science, 391, 1316-1317; and Cheng, M. et al. (2026). Sycophantic AI decreases prosocial intentions and promotes dependence. Science, 391, eaec8352. We argue that AI sycophancy reflects something deeper than a training problem: it reflects the consequences of positioning AI within a relational role it cannot genuinely occupy. Drawing on Dweck's unified theory of psychological needs, we distinguish between needs that require genuine relational presence (acceptance, trust) and needs that can be supported through structured tools (competence, autonomy). Agentic AI systems organized around skill-building and structured reasoning are less susceptible to sycophancy dynamics because the interaction does not rest on emotional affirmation.
Refoua et al. (Wed,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: