Adaptive History Biases Result from Confidence-Weighted Accumulation of past Choices

Anke Braun, Anne E Urai, Tobias H. Donner

TL;DR

This study reveals that our decisions are not just based on current information but are adaptively influenced by past choices, particularly when we're confident about them. By manipulating the predictability of stimuli, researchers found that people flexibly adjust their "history biases" (tendency to repeat or alternate past choices) based on environmental statistics, and this adjustment is crucial for optimal performance, even rivaling the importance of basic sensory perception.

What Problem Does This Paper Address?

We often exhibit "choice history biases," meaning our current decision is influenced by what we decided in previous trials – we tend to repeat or alternate our past choices more than chance would predict. While computational models suggest this arises from accumulating internal decision signals over time, it's unclear whether these signals are tied to the motor action of choosing or are more abstract, "action-independent" representations. Furthermore, it's unknown if these biases can be flexibly adapted to different environments, especially when external feedback is absent, and if our confidence in past decisions plays a role in this adaptation. This paper aims to disentangle these factors, investigating the origin, adaptability, and confidence-dependence of choice history biases in perceptual decision-making.

Methods

The researchers used a challenging visual motion discrimination task (judging if dots moved up or down) with human participants.

  • Experiment 1: They decoupled perceptual choices from motor responses on a trial-by-trial basis. This allowed them to test whether history biases stemmed from the perceptual decision itself or the physical act of pressing a button.
  • Experiment 2: Participants performed the task in environments with varying levels of predictability (repetition probabilities) between successive trials, without receiving external feedback on their accuracy. This setup allowed them to study how observers adapt their history biases to these different environmental statistics (Repetitive, Neutral, Alternating).
  • Modeling: They employed logistic regression models to quantify the influence of past stimuli, choices, and motor responses on current choices. They also developed models to assess how confidence, proxied by choice accuracy and reaction time, modulated the adjustment of history biases.

Datasets

Not discussed in the paper. The study used data collected from human participants performing the described experiments.

Key Insights and Ideas

  • Choice, Not Motor Action, Drives Bias: History biases are primarily driven by previous perceptual choices, not the motor responses used to report those choices. This suggests the underlying decision signals are abstract and action-independent.
  • Flexible Adaptation to Environment: Humans can dynamically adjust their choice history biases to match the statistical regularities of their environment, even without external feedback. They can learn to repeat choices in predictable environments and alternate them in environments where alternation is beneficial.
  • Confidence is Key: The degree to which individuals adjust their history biases is significantly modulated by their confidence in past decisions. More confident past choices have a stronger influence on future bias adjustment.
  • Bias Adjustment is as Important as Sensitivity: The ability to adapt history biases to the environment can be as critical for optimizing performance as an individual's basic perceptual sensitivity to the current stimulus.

Key Results

  • In Experiment 1, previous choices had a significant impact on current choices, while previous motor responses did not.
  • In Experiment 2, participants showed significant adjustments in their history biases, shifting towards repetition in "Repetitive" environments and alternation in "Alternating" environments, compared to a "Neutral" baseline.
  • The "adaptivity index" (measuring how well individual history biases matched environmental templates) was a strong predictor of performance, especially in the biased environments (Repetitive and Alternating).
  • Confidence, measured by both accuracy and reaction time, played a crucial role: the adjustment of history biases was stronger for correct and faster previous choices. For instance, the influence of previous correct choices on bias adjustment was significantly higher than for incorrect choices.

Limitations and Future Work

  • Neural Basis: While the study provides strong psychophysical evidence, the underlying neural mechanisms of this confidence-weighted accumulation process remain to be elucidated. The authors suggest this work sets the stage for future neurophysiological investigations.
  • Generalizability: The study focused on a specific perceptual decision task. Future work could explore if these findings generalize to other decision-making domains, such as economic or social decisions.
  • Confidence Proxies: The study used accuracy and reaction time as proxies for confidence. While these are well-established, exploring other potential neural or behavioral markers of confidence could further refine the understanding.
  • Model Complexity: The authors acknowledge that their confidence model is a simplification and that more complex models might capture additional nuances of decision confidence.

The paper concludes that human observers accumulate action-independent, confidence-modulated decision variables across trials in a flexible manner, allowing them to adapt their behavior to environmental statistics, which is crucial for optimizing performance.

Summary updated

Abstract

Perceptual decision-making is biased by previous events, including the history of preceding choices: observers tend to repeat (or alternate) their judgments of the sensory environment more often than expected by chance. Computational models postulate that these so-called choice history biases result from the accumulation of internal decision signals across trials. Here, we provide psychophysical evidence for such a mechanism and its adaptive utility. Male and female human observers performed different variants of a challenging visual motion discrimination task near psychophysical threshold. In a first experiment, we decoupled categorical perceptual choices and motor responses on a trial-by-trial basis. Choice history bias was explained by previous perceptual choices, not motor responses, highlighting the importance of internal decision signals in action-independent formats. In a second experiment, observers performed the task in stimulus environments containing different levels of autocorrelation and providing no external feedback about choice correctness. Despite performing under overall high levels of uncertainty, observers adjusted both the strength and the sign of their choice history biases to these environments. When stimulus sequences were dominated by either repetitions or alternations, the individual degree of this adjustment of history bias was about as good a predictor of individual performance as individual perceptual sensitivity. The history bias adjustment scaled with two proxies for observers' confidence about their previous choices (accuracy and reaction time). Together, our results are consistent with the idea that action-independent, confidence-modulated decision variables are accumulated across choices in a flexible manner that depends on decision-makers' model of their environment.<b>SIGNIFICANCE STATEMENT</b> Decisions based on sensory input are often influenced by the history of one's preceding choices, manifesting as a bias to systematically repeat (or alternate) choices. We here provide support for the idea that such choice history biases arise from the context-dependent accumulation of a quantity referred to as the decision variable: the variable's sign dictates the choice and its magnitude the confidence about choice correctness. We show that choices are accumulated in an action-independent format and a context-dependent manner, weighted by the confidence about their correctness. This confidence-weighted accumulation of choices enables decision-makers to flexibly adjust their behavior to different sensory environments. The bias adjustment can be as important for optimizing performance as one's sensitivity to the momentary sensory input.
BibTeX
@article{braun2018adaptive,
  author = {Anke Braun and Anne E Urai and Tobias H. Donner},
  title = {{Adaptive History Biases Result from Confidence-Weighted Accumulation of past Choices}},
  journal = {Journal of Neuroscience},
  year = {2018},
  doi = {10.1523/jneurosci.2189-17.2017},
}
Figure