“Toward a psychology of real-life decision-making”
- Novel developments in biologging technology allow for precise measurement of individual-level behavior across time and space. Additionally, novel technologies allow researchers to quantify in detail the environments in which people are embedded when making decisions.
- Such devices, however, are only useful to cognitive scientists when integrated with formal theoretical frameworks for decision-making. This review outlines a cognitive–computational framework for precise inference about cognitive processes during real-life decision-making.
Understanding human decision-making processes in everyday life is a central, yet rarely addressed, challenge in psychology. Either real-life complexity is reduced by isolating specific aspects of decision-making in highly constrained experimental settings, yielding insights into specific cognitive mechanisms under idealized conditions, or decision-making is studied in real-life contexts, using high-level descriptions of behavior that do not afford fine-grained, process-level insights.
Bridging this gap poses a challenge of both measurement and inference. Recent advances in high-resolution tracking technologies provide novel solutions to many measurement challenges but are rarely integrated with formal psychological theory.
In this article, we review tracking technologies and statistical tools, proposing a cognitive–computational framework that uses high-resolution spatiotemporal data to investigate the mechanisms of real-life decision-making in a theory-driven manner.
Novel methodological tools are emerging to quantify the four observable building blocks of the cognitive–computational framework.
(A) Behavioral constraints, such as reachability and visibility of locations, and energy costs, can be quantified using, for example, geographic information systems (GIS), OpenStreetMap data, or wearable physiological sensors.
(B) Available information, such as personal, social, and environmental, can be quantified using, for example, digital communication, mobile sensing, wearable eye-tracking, field-of-view cameras or video recordings.
(C) Behavioral traces (i.e., changes in a behavioral variable x across time t such as changes in movement speed, physiology or likes on social media) can be measured using, for example, GPS recordings, video data, or digital traces.
(D) Environmental variables that individuals face when making decisions (e.g., autocorrelations in resources, urbanization, or differing degrees of spatiotemporal variability in resources) can be measured using, for example, aerial imaging or panoramic street view data. Different technological tools can be used to quantify different building blocks depending on the context and aim of the study.
High-resolution tracking data can help infer the latent processes of ‘construction’, ‘selection’, ‘transformation’, ‘emission’, and ‘adaptation’.
(A) Data on behavioral constraints can be used to identify the options (circles and rectangles) that are available to a decision-maker (e.g., because they are both reachable and visible).
(B) Selecting information requires individuals to focus on relevant pieces of information among all observable information. Transforming information requires latent value computations for options as a function of the information available about these options. By quantifying, for example, how many individuals chose an option (e.g., crossing the street at a red light), the value of this option, Vt, can be transformed. This transformation can occur through mechanisms of varying computational complexity, such as heuristic processes, temporal updating, or evidence accumulation.
(C) The emission process, linking decision-making to observable behavioral traces, can be inferred using deterministic (e.g., decision trees and manual annotation) or probabilistic methods (e.g., state–space models). Decision trees can be used to identify decision points, Dt, based on predefined cutoffs (e.g., a decision to cross the street is made when the change in movement speed across time t exceeds ε). When no clear cutoffs exist but the relation between behavioral indicators and decision points is more complex, manual annotation of, for example, video data, can be used to identify decisions. State–space models provide a probabilistic account of emission processes, allowing for the statistical estimation of decision points as switches between distinct behavioral states across time.
(D) By estimating selection and transformation processes across environments, researchers can gain insights into how decision-making is adapted to different environments.
Cognitive and behavioral scientists have long been concerned that a gap between the laboratory and real life makes it difficult to apply insights from constrained laboratory experiments to complex real-life behavior.
The mismatch poses significant challenges for practitioners and applied scientists aiming to infer the causes of behavior observed outside the laboratory.
Studying the mechanisms of decision-making outside the laboratory requires robust measurement of behavioral and environmental variables, as well as tools to infer latent quantities and processes.
This review summarizes novel technological developments and statistical tools that enable researchers to study behavior in natural environments with ever-increasing precision and richness.
On the one hand, these advances offer insights into the mechanisms underlying behavior in real life.
On the other hand, theoretically grounded approaches are needed to leverage such data and facilitate the transfer of knowledge across domains.
Our review aligns with recent calls to integrate high-resolution data with cognitive modeling in computational psychiatry, particularly via reinforcement learning models applied to digital data.
The present review shows how such an approach can inform specific processes of decision-making across a broader range of real-life environments and modeling frameworks, illustrating how formal models of decision-making offer a common language to achieve commensurability between laboratory and real-life findings. Importantly, precise knowledge of the cognitive mechanisms underlying decision-making outside the laboratory is a prerequisite not only for designing effective interventions to improve mental health and well-being but also for helping individuals adaptively navigate their everyday environments.
Surtil : 