Talluri, B.C., Urai, A.E., Bronfman, Z.Z., Brezis, N., Tsetsos, K., Usher, M., Donner, T.H. (2021) Choices change the temporal weighting of decision evidence, Journal of Neurophysiology, 125(4), 1468-1481 (2021).
By Bharath Chandra Talluri & Tobias Donner.
A prominent idea in perceptual decision-making research is that evidence about the state of the environment is continuously accumulated across time (Gold and Shadlen, 2007). In laboratory experimental tasks with time-varying evidence (such as a random dot motion stimulus) accuracy in judgments can be maximised by weighing the evidence equally across time (Bogacz et al., 2006). This strategy gives rise to what is commonly referred to as the “flat temporal weighting profile”. Yet, human and non-human subjects deviate from this performance- maximising strategy by giving more weight to evidence early in time, a phenomenon called “primacy”, or more weight to evidence later in time, a phenomenon called “recency”. The factors governing the differences in these distinct weighting profiles are largely unknown but identifying them may provide insights into the different mechanisms underlying decision-making. In this new study, we sought to investigate if intermittent categorical choices are one such factor. To this end, we analysed data from two previously published studies (Bronfman et al., 2015; Talluri et al., 2018) using a similar task design but with two different stimulus modalities: perceptual (random dot motion), and symbolic (numerical integration).
Human participants estimated the mean across two intervals of perceptual or numerical evidence, as a continuous judgment. After the first interval, subjects received a cue prompting them to either report a categorical choice about the first interval, or a simple button press that is independent of the preceding evidence. The cue was randomized across trials so that it is difficult for subjects to guess the type of intermittent response at the beginning of each trial. We can thus compare the weights subjects gave to each interval, depending on whether they reported a categorical choice midway through the trial or not. In both the datasets, when subjects made a choice-independent button press, their weights were higher for the second interval. This pattern however flipped, when the intermittent response was a categorical choice report, with higher weights for the evidence informing the choice. This change in relative weighting between early and late evidence was consistent across participants in both datasets. Thus, making an intermittent categorical choice flipped the participants’ weighting from recency to primacy. But what could have caused this flip?
The sum of weights across the two intervals was similar between the Choice and No-Choice conditions, a pattern observed in both datasets. Surprisingly, the difference in weights between the two conditions was negatively correlated across intervals. This combination of results suggests that an intermittent categorical choice increased the weight to the preceding evidence, but at the cost of reducing the weight to subsequent evidence. The pattern of results is consistent with the idea that categorical choices change the state of the decision-making regions in the brain. Such a state change can be mediated by transient neuromodulatory input from the central arousal systems in the brain. Several studies from our lab and others showed that non-luminance mediated pupil dilations are a marker of the responses of these arousal systems (de Gee et al., 2017, 2020). We hence analysed pupil data from participants performing the perceptual task. We found that the pupil dilation following the intermittent response was higher for Choice condition compared to No-Choice condition, reflecting the transient activation of arousal systems following choice-commitment. Across the group, participants with a higher average pupil response following the categorical choice had lower weights to the second interval while no such relation was found on No-Choice condition. These results agree with the idea that choice commitment triggers a state change mediated by the central arousal systems.
In our previous work, we showed that subjects show a selective decrease in sensitivity to choice-inconsistent evidence in the second interval, a signature of confirmation bias in our task (Talluri et al., 2018). We then asked whether the confirmation bias identified before, and the non-selective reduction in sensitivity identified in this work are related. These two biases could be independent, suggesting mediation by distinct mechanisms, or tightly correlated across subjects suggesting a common underlying mechanism. We found the latter to be the case, where subjects with a greater choice-commitment bias also showed a stronger confirmation bias.
To summarise, we found that categorical decisions midway through the trial change the relative weighting of early and late evidence. This finding was established in two independent datasets with perceptual and numerical judgments. Temporal-weighting profiles can thus be flexibly altered within the same individual and task by asking for an intermittent choice. This suggests that the dynamics of evidence-weighting in decision-making is context-dependent. It is tempting to speculate that the overall reduction in sensitivity following a choice-commitment results from an actively induced state change in the decision-making circuits of the brain: Under settings where a subject is uncertain about the state of the environment, a neuromodulatory input induced by the choice can push decision-circuits into a stable state corresponding to choice commitment. Future circuit modeling and neurophysiology work should test these ideas.
Simple summary of our Confirmation bias paper.
Bogacz, R., Brown, E., Moehlis, J., Holmes, P., Cohen, J.D. (2006) The physics of optimal decision making: A formal analysis of models of performance in two-alternative forced-choice tasks. Psychol Rev 113:700–765.
Bronfman, Z.Z., Brezis, N., Moran, R., Tsetsos, K., Donner, T.H., Usher, M. (2015) Decisions reduce sensitivity to subsequent information. Proc R Soc B Biol Sci 282:20150228.
de Gee, J.W., Colizoli, O., Kloosterman, N.A., Knapen, T., Nieuwenhuis, S., Donner, T.H. (2017) Dynamic modulation of decision biases by brainstem arousal systems. eLife 6:e23232.
de Gee. J.W., Tsetsos. K., Schwabe. L., Urai. A.E., McCormick. D., McGinley. M.J., Donner. T.H. (2020) Pupil-linked phasic arousal predicts a reduction of choice bias across species and decision domains Gold JI, Grüschow M, Ebitz RB, eds. eLife 9:e54014.
Gold, J.I., Shadlen, M.N. (2007) The Neural Basis of Decision Making. Annu Rev Neurosci 30:535–574.
Talluri, B.C., Urai, A.E., Tsetsos, K., Usher, M., Donner, T.H. (2018) Confirmation Bias through Selective Overweighting of Choice-Consistent Evidence. Curr Biol 28:3128-3135.e8.