An Affective-Taxis Hypothesis for Alignment and Interpretability
Eli Sennesh; Maxwell Ramstead · 2025 · AGI 2025 (Springer LNAI), pp. 188-201 background low priority coded
Main argument
Thesis: an 'affectivist' reframing of alignment - goals and values reconceived as AFFECTIVE TAXIS (valence-guided navigation), with affective valence explained via evolutionary-developmental and computational neuroscience (free energy principle / active inference); proposes a computational model of affect based on taxis navigation with evidence from a tractable model organism, and discusses its role in alignment.
Why it matters here
The affectivist counterpoint: grounds goals/values in affective valence (taxis navigation, active inference) rather than utility or norms. Marginal to the dissertation except as the naturalistic-embodiment contrast to Noller's phronesis/affect argument - what it would take to give machines the affective grounding Noller says virtue requires.
Reading notes
Compact treatment (14pp; computational-neuroscience framing). Abstract read.
Sennesh, E., & Ramstead, M. (2025). An Affective-Taxis Hypothesis for Alignment and Interpretability. In AGI 2025 Proceedings Part II. Springer.