Choice History Bias in Perceptual Decision Making

This direction is split into other directions: choice history bias. Its last version stays readable and citable here.

Version 1 of 131 papers38 conceptsfrozen 2026-10-04

In classical models of decision-making, each trial is often treated as an independent event. An agent receives sensory evidence, processes it, and makes a choice. However, a growing body of research across neuroscience, psychology, and multi-agent systems reveals that biological and artificial agents are rarely so "forgetful." Instead, decisions are heavily influenced by choice history bias—a phenomenon where previous actions and their outcomes dictate current behavior, often competing with or even overriding immediate sensory data.

The Mechanism of Behavioral Inertia

At its core, choice history bias functions as a form of "behavioral inertia." Rather than starting each decision from a neutral state, the brain maintains a running internal prior based on recent experience. Charles Findling and Félix Hubert have shown that these subjective priors are often shaped more by an agent's own recent choices than by the actual statistical frequency of external events (Brain-wide representations of prior information in mouse decision-making).

This bias manifests in two primary ways:

  1. Perseveration: The tendency to repeat a previous action, regardless of the outcome.
  2. Reinforcement: The tendency to repeat actions that were previously rewarded.

In tasks where sensory information is ambiguous or difficult to process, agents default to these internal heuristics. Dario Campagner and colleagues demonstrated a "double dissociation" in tactile tasks: while correct choices are driven by sensory input, errors are almost entirely predicted by the memory of previous successful choices (Prediction of Choice From Competing Mechanosensory and Choice-Memory Cues During Active Tactile Decision Making).

Explaining "Lapses" as Deterministic Bias

One of the most significant insights from this research direction is the reinterpretation of "lapses"—errors made on seemingly easy trials. Traditionally dismissed as random noise or inattention, researchers like Nicholas A. Roy and Jonathan W. Pillow argue that these errors are often the predictable result of over-weighting trial history (Efficient inference for time-varying behavior during learning).

Diksha Gupta and Carlos D. Brody further explain this using the framework of evidence accumulation. They suggest that history biases shift the "starting point" of the decision process. If the bias is strong enough, the agent may reach a decision threshold before any sensory evidence is even processed, leading to fast, history-driven errors (Trial-history biases in evidence accumulation can give rise to apparent lapses in decision-making).

The Neural and Computational Basis

This bias is not localized to a single "decision center" but is represented across nearly 30% of the brain, from sensory to motor regions (Brain-wide representations of prior information in mouse decision-making). It is also subject to neurochemical modulation; for instance, Jan Willem de Gee found that increasing catecholamine levels (dopamine and noradrenaline) reduces choice history bias, forcing the brain to rely more on current sensory evidence (Catecholamines reduce choice history biases in perceptual decision making).

In multi-agent systems, such as the Minority Game, this reliance on history is modeled through "strategy scores." Agents track the hypothetical success of various strategies over time and select the one with the highest cumulative performance (The minority game with variable payoffs). While this allows for self-organization and coordination without central control, it also highlights a fundamental trade-off: a system that relies too heavily on history may become rigid and unable to adapt to new information.

Why It Works: Rationality in a Changing World

While these biases often lead to errors in controlled laboratory settings, they may be "rational" in the wild. André G. Mendonça and Alexandre Pouget suggest that trial-by-trial biases are signatures of a brain that assumes the world is constantly changing (The impact of learning on perceptual decisions and its implication for speed-accuracy tradeoffs). In a non-stationary environment, your most recent experience is often the best predictor of the immediate future. Choice history bias, therefore, is not a "bug" of biological hardware, but a feature of a system optimized for continuous learning and adaptation.

Papers in this version

History & citationVersion 1 · 2026-10-04

Split into other directions · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_choice_history_bias_in_perceptual_decisi_b7334dbc,
  title = {Choice History Bias in Perceptual Decision Making},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/choice-history-bias-in-perceptual-decision-making/rdr_b7334dbcc0db40008aa39c5a84aab35f},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-20},
}

This exact version (version 1, frozen text):

@misc{lacuna_choice_history_bias_in_perceptual_decisi_b7334dbc_v1,
  title = {Choice History Bias in Perceptual Decision Making},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/choice-history-bias-in-perceptual-decision-making/rdr_b7334dbcc0db40008aa39c5a84aab35f/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-20},
}