Amelia R. Hunt

Choices in a focus-divide dilemma tend to be variable and suboptimal. How readily will people adopt a simple, rational choice rule to improve performance? We asked participants to choose where to stand to throw a beanbag into one of two possible hoops (Experiments 1A, 1B, 1C), and to choose where to fixate to discriminate a target that could appear in one of two possible boxes (Experiment 2). The rational choice follows a simple logic: when hoops/boxes are close together, standing/fixating near their midpoint is optimal. When hoops/boxes are far apart, standing at/fixating one hoop/box will optimise success. We attempted to nudge participants towards optimal choices by providing circumstances in which they carried out the correct solution, and then returning them to the original context to freely choose. In both experiments, participants executed optimal choices during the intervention, and then immediately returned to a sub-optimal pattern of responding when presented with the original context. Experiments 3A and 3B demonstrate that the rational decision rule can be easily applied and improves performance. We conclude that variable and idiosyncratic responses are a persistent choice strategy. Borrowing concepts from classic animal learning studies, we suggest rule-following risks missing opportunities for reinforcement learning afforded by choice variability.

How do we decide where to search for a target? Optimal search relies on first considering the relative informational value of different locations and then executing eye movements to the best options. However, many participants consistently move their eyes to locations that can be easily ascertained to neither contain the target nor provide new information about the target’s location. Here, we asked whether this suboptimal search behaviour represents a specific example of a general tendency towards precrastination: starting sub-goals of a task before they are needed, and in so doing, spending longer time on doing the task than is necessary. To test this hypothesis, we asked 200 participants to do two tasks: retrieve two heavy buckets (one close and one far) and search for a line segment. Precrastination is defined as consistently picking up the closer bucket first, versus the more efficient strategy of picking up the farther bucket first. Search efficiency is the proportion of fixations directed to more cluttered regions of the search array. Based on the pilot data, we predicted an association of precrastination with inefficient search strategies. Personality inventories were also administered to identify stable characteristics associated with these strategies. In the final dataset, there was no clear association between search strategy and precrastination, nor did these correlate strongly with any of the personality measures collected. This article received in-principle acceptance (IPA) at Royal Society Open Science on 29 January 2020. The accepted Stage 1 version of the manuscript, not including results and discussion, may be found at https://osf.io/p2sjx . This preregistration was performed prior to data collection and analysis.

All measures of behaviour have a temporal context. Changes in behaviour over time often take a similar form: monotonically decreasing or increasing towards an asymptote. Whether these behavioural dynamics are the object of study or a nuisance variable, their inclusion in models of data makes conclusions more complete, robust and well-specified, and can contribute to theory development. Here we demonstrate that asymptotic regression is a relatively simple tool that can be applied to repeated-measures data to estimate three parameters: starting point, rate of change, and asymptote. Each of these parameters has a meaningful interpretation in terms of ecological validity, behavioural dynamics and performance limits, respectively. They can also be used to help decide how many trials to include in an experiment, and as a principled approach to reducing noise in data. We demonstrate the broad utility of asymptotic regression for modelling the effect of the passage of time within a single trial, and for changes over trials of an experiment, using three examples. An important limit of asymptotic regression is that it cannot be applied to data that is stationary or changes non-monotonically. But for data that has performance changes that progress steadily towards an asymptote, as many behavioural measures do, it is a simple and powerful tool for describing those changes.

Does our conception of a target affect how we look for it? Previous investigations have shown that search for an object with a unique identity is more efficient than search based on unique orientation. Here we asked whether giving an orientation target an identity label is sufficient to boost search efficiency. Participants completed two blocks of search for a line tilted 45 ̊ to the right. Between blocks we induced an identity-based connection between the participant and search target using the self-reference effect (SRE): participants learned to associate the target to the label “you” and distractor lines to the label “strangers”. A second experiment used the same paradigm but introduced a more intuitive relationship between the label and the object form by embedding the line in a circle and adding two dots to create face pictograms. The results of both experiments replicated the standard SRE as measured by reaction time, but search was similarly inefficient when the instructions were to search “for you” as they were in search for orientation. The results suggest search strategies are not strongly influenced by how a target is conceived, ruling out an intuitively appealing explanation for the large individual differences typically observed in search performance.

Prioritising different objectives is a common dilemma of daily life. A simple and effective decision rule is to focus resources when the tasks are difficult, and divide when tasks are easy. Nonetheless, in experimental paradigms of this dilemma, participants make highly variable and suboptimal strategic decisions when asked to allocate resources to two competing goals that vary in difficulty. We developed a virtual dilemma in which participants had to choose where to park a fire truck between houses of varying distances apart. Unlike in the previous versions of the dilemma, participants approached the optimal strategy in this task. Three key differences between the virtual version and previous versions of the task were investigated: 1) Task framing (how concrete and important the objectives are); 2) Agency (not only making strategic choices about how to prioritise tasks, but also carrying out those tasks) and 3) Uncertainty (in the virtual version choice outcomes were more certain and more closely resembled a step function, while in previous versions success on each trial was a probability curve decreasing with difficulty). We found that making tasks more abstract and adding control over task execution to our virtual focus-divide dilemma did not worsen strategic decisions, suggesting framing and agency could not explain why participants are better at the virtual version. When adding variability to outcomes, however, decisions shifted away from optimal. The results suggest choices become more variable when the outcome is less certain, consistent with exploration of response alternatives triggered by an inability to predict success.

Foraging refers to search involving multiple targets or multiple types of targets, and as a model task has a long history in animal behaviour and human cognition research. Foraging behaviour is usually operationalized using summary statistics, such as average distance covered during target collection (the path length) and the frequency of switching between target types. We recently introduced an alternative approach, which is to model each instance of target selection as random selection without replacement. Our model produces estimates of a set of foraging biases, such as a bias to select closer targets or targets of a particular category. Here we apply this model to predict individual target selection events. We add a new start position bias to the model, and generate foraging paths using the parameters estimated from individual participants’ pre-existing data. The model predicts which target the participant will select next with a range of accuracy from 43% to 69% across participants (chance is 11%). The model therefore explains a substantial proportion of foraging behaviour in this paradigm. The situations where the model makes errors reveal useful information to guide future research on those aspects of foraging that we have not yet explained.

Foraging refers to search involving multiple targets or multiple types of targets, and as a model task has a long history in animal behaviour and human cognition research. Foraging behaviour is usually operationalized using summary statistics, such as average distance covered during target collection (the path length) and the frequency of switching between target types. We recently introduced an alternative approach, which is to model each instance of target selection as random selection without replacement. Our model produces estimates of a set of foraging biases, such as a bias to select closer targets or targets of a particular category. Here we apply this model to predict individual target selection events. We add a new start position bias to the model, and generate foraging paths using the parameters estimated from individual participants’ pre-existing data. The model predicts which target the participant will select next with a range of accuracy from 43\% to 69\% across participants (chance is 11\%). The model therefore explains a substantial proportion of foraging behaviour in this paradigm. The situations where the model makes errors reveal useful information to guide future research on those aspects of foraging that we have not yet explained.

Foraging entails finding multiple targets sequentially. In humans and other animals, a key observation has been a tendency to forage in ‘runs’ of the same target type. This tendency is context-sensitive, and in humans, it is strongest when the targets are difficult to distinguish from the distractors. Many important questions have yet to be addressed about this and other tendencies in human foraging, and a key limitation is a lack of precise measures of foraging behaviour. The standard measures tend to be run statistics, such as the maximum run length and the number of runs. But these measures are not only interdependent, they are also constrained by the number and distribution of targets, making it difficult to make inferences about the effects of these aspects of the environment on foraging. Moreover, run statistics are underspecified about the underlying cognitive processes determining foraging behaviour. We present an alternative approach: modelling foraging as a procedure of generative sampling without replacement, implemented in a Bayesian multilevel model. This allows us to break behaviour down into a number of biases that influence target selection, such as the proximity of targets and a bias for selecting targets in runs, in a way that is not dependent on the number of targets present. Our method thereby facilitates direct comparison of specific foraging tendencies between search environments that differ in theoretically important dimensions. We demonstrate the use of our model with simulation examples and re-analysis of existing data. We believe our model will provide deeper insights into visual foraging and provide a foundation for further modelling work in this area.

Foraging entails finding multiple targets sequentially. In humans and other animals, a key observation has been a tendency to forage in `runs' of the same target type. This tendency is context-sensitive, and in humans, it is strongest when the targets are difficult to distinguish from the distractors. Many important questions have yet to be addressed about this and other tendencies in human foraging, and a key limitation is a lack of precise measures of foraging behaviour. The standard measures tend to be run statistics, such as the maximum run length and the number of runs. But these measures are not only interdependent, they are also constrained by the number and distribution of targets, confounding any inferences about the effects of these aspects of the environment on foraging. Moreover, run statistics are underspecified about the underlying cognitive processes determining foraging behaviour. We present an alternative approach: modelling foraging as a procedure of generative sampling without replacement, implemented in a Bayesian multilevel model. This allows us to break behaviour down into a number of biases that influence target selection, such as the proximity of targets and a bias for selecting targets in runs, in a way that is not dependent on the number of targets present. Our method thereby facilitates direct comparison of specific foraging tendencies between search environments that differ in theoretically important dimensions. We demonstrate the use of our model with simulation examples and re-analysis of existing data. We believe our model will provide deeper insights into visual foraging and provide a foundation for further modelling work in this area.

We measured the efficiency of 30 participants as they searched through simple line segment stimuli and through a set of complex icons. We observed a dramatic shift from highly variable, and mostly inefficient, strategies with the line segments, to uniformly efficient search behaviour with the icons. These results demonstrate that changing what may initially appear to be irrelevant, surface-level details of the task can lead to large changes in measured behaviour, and that visual primitives are not always representative of more complex objects.