Overcoming indecision by changing the decision boundary

The neural network RTNet exhibits the signatures of human perceptual decision-making

Convolutional neural networks show promise as models of biological vision. However, their decision behaviour, including the facts that they are deterministic and use equal numbers of computations for easy and difficult stimuli, differs markedly from human decision-making, thus limiting their applicability as models of human perceptual behaviour. Here we develop a new neural network, RTNet, that generates stochastic decisions and human-like response time (RT) distributions. We further performed comprehensive tests that showed RTNet reproduces all foundational features of human accuracy, RT and confidence and does so better than all current alternatives. To test RTNet's ability to predict human behaviour on novel images, we collected accuracy, RT and confidence data from 60 human participants performing a digit discrimination task. We found that the accuracy, RT and confidence produced by RTNet for individual novel images correlated with the same quantities produced by human participants. Critically, human participants who were more similar to the average human performance were also found to be closer to RTNet's predictions, suggesting that RTNet successfully captured average human behaviour. Overall, RTNet is a promising model of human RTs that exhibits the critical signatures of perceptual decision-making.

Normative evidence weighting and accumulation in correlated environments

The brain forms certain deliberative decisions following normative principles related to how sensory observations are weighed and accumulated over time. Previously we showed that these principles can account for how people adapt their decisions to the temporal dynamics of the observations (Glaze et al., 2015). Here we show that this adaptability extends to accounting for correlations in the observations, which can have a dramatic impact on the weight of evidence provided by those observations. We tested online human participants on a novel visual-discrimination task with pairwise-correlated observations. With minimal training, the participants adapted to uncued, trial-by-trial changes in the correlations and produced decisions based on an approximately normative weighting and accumulation of evidence. The results highlight the robustness of our brain's ability to process sensory observations with respect to not just their physical features but also the weight of evidence they provide for a given decision.

Expressions for Bayesian confidence of drift diffusion observers in fluctuating stimuli tasks

We introduce a new approach to modelling decision confidence, with the aim of enabling computationally cheap predictions while taking into account, and thereby exploiting, trial-by-trial variability in stochastically fluctuating stimuli. Using the framework of the drift diffusion model of decision making, along with time-dependent thresholds and the idea of a Bayesian confidence readout, we derive expressions for the probability distribution over confidence reports. In line with current models of confidence, the derivations allow for the accumulation of "pipeline" evidence that has been received but not processed by the time of response, the effect of drift rate variability, and metacognitive noise. The expressions are valid for stimuli that change over the course of a trial with normally-distributed fluctuations in the evidence they provide. A number of approximations are made to arrive at the final expressions, and we test all approximations via simulation. The derived expressions contain only a small number of standard functions, and require evaluating only once per trial, making trial-by-trial modelling of confidence data in stochastically fluctuating stimuli tasks more feasible. We conclude by using the expressions to gain insight into the confidence of optimal observers, and empirically observed patterns.

Multiphasic value biases in fast-paced decisions

Perceptual decisions are biased toward higher-value options when overall gains can be improved. When stimuli demand immediate reactions, the neurophysiological decision process dynamically evolves through distinct phases of growing anticipation, detection, and discrimination, but how value biases are exerted through these phases remains unknown. Here, by parsing motor preparation dynamics in human electrophysiology, we uncovered a multiphasic pattern of countervailing biases operating in speeded decisions. Anticipatory preparation of higher-value actions began earlier, conferring a 'starting point' advantage at stimulus onset, but the delayed preparation of lower-value actions was steeper, conferring a value-opposed buildup-rate bias. This, in turn, was countered by a transient deflection toward the higher-value action evoked by stimulus detection. A neurally-constrained process model featuring anticipatory urgency, biased detection, and accumulation of growing stimulus-discriminating evidence, successfully captured both behavior and motor preparation dynamics. Thus, an intricate interplay of distinct biasing mechanisms serves to prioritise time-constrained perceptual decisions.

Using cognitive modeling to examine the effects of competition on strategy and effort in races and tournaments

We investigated the effects of two types of competition, races and tournaments (as well as an individual challenge and a do-your-best condition), on two different aspects of performance: effort and strategy. In our experiment, 100 undergraduate participants completed a simple cognitive task under four experimental conditions (in a repeated-measures design) based on different types of competitions and challenges. We used the Linear Ballistic Accumulator to quantify the effects of competition on strategy and effort. The results reveal that competition produced changes in strategy rather than effort, and that trait competitiveness had minimal impact on how people responded to competition. This suggests individuals are more likely to adjust their strategy in competitions, and the uncertainty created by different competition types influences the direction of these strategy adjustments.

Normative decision rules in changing environments

Models based on normative principles have played a major role in our understanding of how the brain forms decisions. However, these models have typically been derived for simple, stable conditions, and their relevance to decisions formed under more naturalistic, dynamic conditions is unclear. We previously derived a normative decision model in which evidence accumulation is adapted to fluctuations in the evidence-generating process that occur during a single decision (Glaze et al., 2015), but the evolution of commitment rules (e.g. thresholds on the accumulated evidence) under dynamic conditions is not fully understood. Here, we derive a normative model for decisions based on changing contexts, which we define as changes in evidence quality or reward, over the course of a single decision. In these cases, performance (reward rate) is maximized using decision thresholds that respond to and even anticipate these changes, in contrast to the static thresholds used in many decision models. We show that these adaptive thresholds exhibit several distinct temporal motifs that depend on the specific predicted and experienced context changes and that adaptive models perform robustly even when implemented imperfectly (noisily). We further show that decision models with adaptive thresholds outperform those with constant or urgency-gated thresholds in accounting for human response times on a task with time-varying evidence quality and average reward. These results further link normative and neural decision-making while expanding our view of both as dynamic, adaptive processes that update and use expectations to govern both deliberation and commitment.

The Cognition/Metacognition Trade-Off

Integration to boundary is an optimal decision algorithm that accumulates evidence until the posterior reaches a decision boundary, resulting in the fastest decisions for a target accuracy. Here, we demonstrated that this advantage incurs a cost in metacognitive accuracy (confidence), generating a cognition/metacognition trade-off. Using computational modeling, we found that integration to a fixed boundary results in less variability in evidence integration and thus reduces metacognitive accuracy, compared with a collapsing-boundary or a random-timer strategy. We examined how decision strategy affects metacognitive accuracy in three cross-domain experiments, in which 102 university students completed a free-response session (evidence terminated by the participant's response) and an interrogation session (fixed number of evidence samples controlled by the experimenter). In both sessions, participants observed a sequence of evidence and reported their choice and confidence. As predicted, the interrogation protocol (preventing integration to boundary) enhanced metacognitive accuracy. We also found that in the free-response sessions, participants integrated evidence to a collapsing boundary-a strategy that achieves an efficient compromise between optimizing choice and metacognitive accuracy.

Trading accuracy for speed over the course of a decision

Humans and other animals often need to balance the desire to gather sensory information (to make the best choice) with the urgency to act, facing a speed-accuracy tradeoff (SAT). Given the ubiquity of SAT across species, extensive research has been devoted to understanding the computational mechanisms allowing its regulation at different timescales, including from one context to another, and from one decision to another. However, animals must frequently change their SAT on even shorter timescales-that is, over the course of an ongoing decision-and little is known about the mechanisms that allow such rapid adaptations. The present study aimed at addressing this issue. Human subjects performed a decision task with changing evidence. In this task, subjects received rewards for correct answers but incurred penalties for mistakes. An increase or a decrease in penalty occurring halfway through the trial promoted rapid SAT shifts, favoring speeded decisions either in the early or in the late stage of the trial. Importantly, these shifts were associated with stage-specific adjustments in the accuracy criterion exploited for committing to a choice. Those subjects who decreased the most their accuracy criterion at a given decision stage exhibited the highest gain in speed, but also the highest cost in terms of performance accuracy at that time. Altogether, the current findings offer a unique extension of previous work, by suggesting that dynamic cha*nges in accuracy criterion allow the regulation of the SAT within the timescale of a single decision.NEW & NOTEWORTHY Extensive research has been devoted to understanding the mechanisms allowing the regulation of the speed-accuracy tradeoff (SAT) from one context to another and from one decision to another. Here, we show that humans can voluntarily change their SAT on even shorter timescales-that is, over the course of a decision. These rapid SAT shifts are associated with dynamic adjustments in the accuracy criterion exploited for committing to a choice.

Children perform extensive information gathering when it is not costly

Humans often face decisions where little is known about the choice options. Gathering information prior to making a choice is an important strategy to improve decision making under uncertainty. This is of particular importance during childhood and adolescence, when knowledge about the world is still limited. To examine how much information youths gather, we asked 107 children (8-9 years, N = 30), early (12-13 years, N = 41) and late adolescents (16-17 years, N = 36) to perform an information sampling task. We find that children gather significantly more information before making a decision compared to adolescents, but only if it does not come with explicit costs. Using computational modelling, we find that this is because children have reduced subjective costs for gathering information. Our findings thus demonstrate how children overcome their limited knowledge and neurocognitive constraints by deploying excessive information gathering, a developmental feature that could inform aberrant information gathering in psychiatric disorders.

Humans sacrifice decision-making for action execution when a demanding control of movement is required

A growing body of evidence suggests that decision-making and action execution are governed by partly overlapping operating principles. Especially, previous work proposed that a shared decision urgency/movement vigor signal, possibly computed in the basal ganglia, coordinates both deliberation and movement durations in a way that maximizes the reward rate. Recent data support one aspect of this hypothesis, indicating that the urgency level at which a decision is made influences the vigor of the movement produced to express this choice. Here we investigated whether, conversely, the motor context in which a movement is executed determines decision speed and accuracy. Twenty human subjects performed a probabilistic decision task in which perceptual choices were expressed by reaching movements toward targets whose size and distance from a starting position varied in distinct blocks of trials. We found strong evidence for an influence of the motor context on most of the subjects' decision policy, but contrary to the predictions of the "shared regulation" hypothesis, we observed that slow movements executed in the most demanding motor blocks in terms of accuracy were often preceded by faster and less accurate decisions compared with blocks of trials in which big targets allowed expression of choices with fast and inaccurate movements. These results suggest that decision-making and motor control are not regulated by one unique "invigoration" signal determining both decision urgency and action vigor, but more likely by independent, yet interacting, decision urgency and movement vigor signals.NEW & NOTEWORTHY Recent hypotheses propose that choices and movements share optimization principles derived from economy, possibly implemented by one unique context-dependent regulation signal determining both processes' speed. In the present behavioral study conducted on human subjects, we demonstrate that action properties indeed influence perceptual decision-making, but that decision duration and action vigor are actually independently set depending on the difficulty of the movement executed to report a choice.

Optimal or not; depends on the task

Decision-making involves a tradeoff between pressures for caution and urgency. Previous research has investigated how well humans optimize this tradeoff, and mostly concluded that people adopt a sub-optimal strategy that over-emphasizes caution. This emphasis reduces how many decisions can be made in a fixed time, which reduces the “reward rate”. However, the strategy that is optimal depends critically on the timing properties of the experiment design: the slower the rate of decision opportunities, the more cautious the optimal strategy. Previous studies have almost uniformly adopted very fast designs, which favor very urgent decision-making. This raises the possibility that previous findings—that humans adopt strategies that are too cautious—could either be ascribed to human caution, or to the experiments’ design. To test this, we used a slowed-down decision-making task in which the optimal strategy was quite cautious. With this task, and in contrast to previous findings, the average strategy adopted across participants was very close to optimal, with about equally many participants adopting too-cautious as too-urgent strategies. Our findings suggest that task design can play a role in inferences about optimality, and that previous conclusions regarding human sub-optimality are conditional on the task settings. This limits claims about human optimality that can be supported by the available evidence.

Optimal models of decision-making in dynamic environments

Nature is in constant flux, so animals must account for changes in their environment when making decisions. How animals learn the timescale of such changes and adapt their decision strategies accordingly is not well understood. Recent psychophysical experiments have shown humans and other animals can achieve near-optimal performance at two alternative forced choice (2AFC) tasks in dynamically changing environments. Characterization of performance requires the derivation and analysis of computational models of optimal decision-making policies on such tasks. We review recent theoretical work in this area, and discuss how models compare with subjects' behavior in tasks where the correct choice or evidence quality changes in dynamic, but predictable, ways.

A theoretical analysis of the reward rate optimality of collapsing decision criteria

A standard assumption of most sequential sampling models is that decision-makers rely on a decision criterion that remains constant throughout the decision process. However, several authors have recently suggested that, in order to maximize reward rates in dynamic environments, decision-makers need to rely on a decision criterion that changes over the course of the decision process. We used dynamic programming and simulation methods to quantify the reward rates obtained by constant and dynamic decision criteria in different environments. We further investigated what influence a decision-maker's uncertainty about the stochastic structure of the environment has on reward rates. Our results show that in most dynamic environments, both types of decision criteria yield similar reward rates, across different levels of uncertainty. This suggests that a static decision criterion might provide a robust default setting.

Integration to boundary in decisions between numerical sequences

Integration-to-boundary is a prominent normative principle used in evidence-based decisions to explain the speed-accuracy trade-off and determine the decision-time. Despite its prominence, however, the decision boundary is not directly observed, but rather is theoretically assumed, and there is still an ongoing debate regarding its form: fixed vs. collapsing. The aim of this study is to show that the integration-to-boundary process extends to decisions between rapid pairs of numerical sequences (2 Hz rate), and to determine the boundary type by directly monitoring the noisy accumulated evidence. In a set of two experiments (supplemented by computational modelling), we demonstrate that integration to a collapsing-boundary takes place in such tasks, ruling out non-integration heuristic strategies. Moreover, we show that participants can adaptively adjust their boundaries in response to reward contingencies. Finally, we discuss the implications to decision optimality and the nature of processes and representations in numerical cognition.

The Urge to Decide and Act: Implications for Brain Function and Dysfunction

Humans and other animals are motivated to act so as to maximize their subjective reward rate. Here, we propose that reward rate maximization is accomplished by adjusting a context-dependent “urgency signal,” which influences both the commitment to a developing action choice and the vigor with which the ensuing action is performed. We review behavioral and neurophysiological data suggesting that urgency is controlled by projections from the basal ganglia to cerebral cortical regions, influencing neural activity related to decision making as well as activity related to action execution. We also review evidence suggesting that different individuals possess specific policies for adjusting their urgency signal to particular contextual variables, such that urgency constitutes an individual trait which jointly influences a wide range of behavioral measures commonly related to the overall quality and hastiness of one’s decisions and actions. Consequently, we argue that a central mechanism for reward rate maximization provides a potential link between personality traits such as impulsivity, as well as some of the motivation-related symptomology of clinical disorders such as depression and Parkinson’s disease.

Time-varying decision boundaries: insights from optimality analysis

The most widely used account of decision-making proposes that people choose between alternatives by accumulating evidence in favor of each alternative until this evidence reaches a decision boundary. It is frequently assumed that this decision boundary stays constant during a decision, depending on the evidence collected but not on time. Recent experimental and theoretical work has challenged this assumption, showing that constant decision boundaries are, in some circumstances, sub-optimal. We introduce a theoretical model that facilitates identification of the optimal decision boundaries under a wide range of conditions. Time-varying optimal decision boundaries for our model are a result only of uncertainty over the difficulty of each trial and do not require decision deadlines or costs associated with collecting evidence, as assumed by previous authors. Furthermore, the shape of optimal decision boundaries depends on the difficulties of different decisions. When some trials are very difficult, optimal boundaries decrease with time, but for tasks that only include a mixture of easy and medium difficulty trials, the optimal boundaries increase or stay constant. We also show how this simple model can be extended to more complex decision-making tasks such as when people have unequal priors or when they can choose to opt out of decisions. The theoretical model presented here provides an important framework to understand how, why, and whether decision boundaries should change over time in experiments on decision-making.

Fast and slow errors: Logistic regression to identify patterns in accuracy-response time relationships

Understanding error and response time patterns is essential for making inferences in several domains of cognitive psychology. Crucial insights on cognitive performance and typical behavioral patterns are disclosed by using distributional analyses such as conditional accuracy functions (CAFs) instead of mean statistics. Several common behavioral error patterns revealed by CAFs are frequently described in the literature: response capture (associated with relatively fast errors), time pressure or urgency paradigms (slow errors), or cue-induced speed–accuracy trade-off (evenly distributed errors). Unfortunately, the standard way of computing CAFs is problematic, because accuracy is averaged in RT bins. Here we present a novel way of analyzing accuracy–RT relationships on the basis of nonlinear logistic regression, to handle these problematic aspects of RT binning. First we evaluate the parametric robustness of the logistic regression CAF through parameter recovery. Second, we apply the function to three existing data sets showing that specific parametric changes in the logistic regression CAF can consistently describe common behavioral patterns (such as response capture, time pressure, and speed–accuracy trade-off). Finally, we discuss potential modifications for future research.

Beta-Blocker Propranolol Modulates Decision Urgency During Sequential Information Gathering

Arbitrating between timely choice and extended information gathering is critical for effective decision making. Aberrant information gathering behavior is thought to be a feature of psychiatric disorders such as schizophrenia and obsessive-compulsive disorder, but we know little about the underlying neurocognitive control mechanisms. In a double-blind, placebo-controlled drug study involving 60 healthy human subjects (30 female), we examined the effects of noradrenaline and dopamine antagonism on information gathering during performance of an information sampling task. We show that modulating noradrenaline function with 40 mg of the β-blocker propranolol leads to decreased information gathering behavior. Modulating dopamine function via a single dose of 400 mg of amisulpride revealed some effects that were intermediate between those of propranolol and placebo. Using a Bayesian computational model, we show that sampling behavior is best explained by inclusion of a nonlinear urgency signal that promotes commitment to an early decision. Noradrenaline blockade promotes the expression of this decision-related urgency signal during information gathering. We discuss the findings with respect to psychopathological conditions that are linked to aberrant information gathering.SIGNIFICANCE STATEMENT Knowing when to stop gathering information and commit to a choice option is nontrivial. This is an important element in arbitrating between information gain and energy conservation. In this double-blind, placebo-controlled drug study, we investigated the role of catecholamines noradrenaline and dopamine on sequential information gathering. We found that blockade of noradrenaline led to a decrease in information gathering. Dopamine blockade showed an intermediate, but nonsignificant, effect. Using a Bayesian computational model, we show that this noradrenaline effect is driven by increased decision urgency, a signal that reflects an escalating subjective cost of sampling. The observation that noradrenaline modulates decision urgency suggests new avenues for treating patients that show information gathering deficits.

Mechanisms Underlying Decision-Making as Revealed by Deep-Brain Stimulation in Patients with Parkinson's Disease

To optimally balance opposing demands of speed and accuracy during decision-making, we must flexibly adapt how much evidence we require before making a choice. Such adjustments in decision thresholds have been linked to the subthalamic nucleus (STN), and therapeutic STN deep-brain stimulation (DBS) has been shown to interfere with this function. Here, we performed continuous as well as closed-loop DBS of the STN while Parkinson’s disease patients performed a perceptual decision-making task. Closed-loop STN DBS allowed temporally patterned STN stimulation and simultaneous recordings of STN activity. This revealed that DBS only affected patients’ ability to adjust decision thresholds if applied in a specific temporally confined time window during deliberation. Only stimulation in that window diminished the normal slowing of response times that occurred on difficult trials when DBS was turned off. Furthermore, DBS eliminated a relative, time-specific increase in STN beta oscillations and compromised its functional relationship with trial-by-trial adjustments in decision thresholds. Together, these results provide causal evidence that the STN is involved in adjusting decision thresholds in distinct, time-limited processing windows during deliberation.

•

We performed temporally patterned stimulation of the subthalamic nucleus in humans

•

During stimulation, Parkinson’s patients performed a perceptual decision-making task

•

Stimulation effects on behavior were confined to a short window during deliberation

•

Here, stimulation affected changes in decision thresholds during difficult decisions

Deep Brain Stimulation of the Subthalamic Nucleus Does Not Affect the Decrease of Decision Threshold during the Choice Process When There Is No Conflict, Time Pressure, or Reward

During a decision process, the evidence supporting alternative options is integrated over time, and the choice is made when the accumulated evidence for one of the options reaches a decision threshold. Humans and animals have an ability to control the decision threshold, that is, the amount of evidence that needs to be gathered to commit to a choice, and it has been proposed that the subthalamic nucleus (STN) is important for this control. Recent behavioral and neurophysiological data suggest that, in some circumstances, the decision threshold decreases with time during choice trials, allowing overcoming of indecision during difficult choices. Here we asked whether this within-trial decrease of the decision threshold is mediated by the STN and if it is affected by disrupting information processing in the STN through deep brain stimulation (DBS). We assessed 13 patients with Parkinson disease receiving bilateral STN DBS six or more months after the surgery, 11 age-matched controls, and 12 young healthy controls. All participants completed a series of decision trials, in which the evidence was presented in discrete time points, which allowed more direct estimation of the decision threshold. The participants differed widely in the slope of their decision threshold, ranging from constant threshold within a trial to steeply decreasing. However, the slope of the decision threshold did not depend on whether STN DBS was switched on or off and did not differ between the patients and controls. Furthermore, there was no difference in accuracy and RT between the patients in the on and off stimulation conditions and healthy controls. Previous studies that have reported modulation of the decision threshold by STN DBS or unilateral subthalamotomy in Parkinson disease have involved either fast decision-making under conflict or time pressure or in anticipation of high reward. Our findings suggest that, in the absence of reward, decision conflict, or time pressure for decision-making, the STN does not play a critical role in modulating the within-trial decrease of decision thresholds during the choice process.