Robert L. Goldstone

The 11 articles in this special issue of Current Directions in Psychological Science examine the converging and diverging paths of human and artificial intelligence (AI) to illuminate the foundations of intelligent behavior. By comparing the mechanisms and limitations of human and machine cognition, these articles reveal how each field can inform the other. Discussed applications of AI to psychology include building computational models of mental processes, automating analyses of rich behavior, automating the generation of materials and procedures for studies in psychology, and providing personal assistants and coaching. Applications of psychology to AI include generating ideas for future AI algorithms by considering how humans reason and applying analysis techniques from psychology to understand AI. Together, the articles advance a broad conception of intelligence that spans humans, machines, and their increasingly intertwined collaborations.

This study explores how scientific simplifications, such as partitioning the brain into distinct regions, shape scientific investigation. One hundred fifty-six psychology undergraduates (novices) and thirty-one neuroscientists (experts) explored a fictional brain segment using neuroimaging and lesioning experiments. Experts engaged in broader exploration than novices, with experimental diversity positively correlating with learning outcomes for both groups. The brain’s partitioning into regions affected participants’ experimental choices and learning success, though these effects varied across individuals. Novices favored testing boundaries between regions, while both groups focused on prototypical “skeletal" locations. These biases affected learning success: novices excelled at identifying signals near boundaries, whereas experts were better at locating signals near the centers of regions. Additionally, both groups demonstrated conceptual expectations associating a discovered signal with only one region and assuming a one-to-one mapping between signals and behavioral tasks. Understanding how simplifications influence scientific experimentation is key to re-engineering the conceptual foundations of science.

STEM education research has historically been under-equipped, relying on standardized tests and questionnaires while other fields deploy space telescopes and particle accelerators. What would happen if we designed research infrastructure for STEM education to address our priority needs? The INTERACT Incubator is a research coordination network whose goal was to develop a field-initiated vision of novel infrastructure that would enable aspirational research that could advance equity in STEM education. The incubator brought together a diverse cohort of experimental social psychologists who study social cues in STEM, experimental cognitive psychologists who study learning in authentic classroom settings, as well as education stakeholders and technologists with expertise in digital infrastructure for education. In Phase 1 we conducted a needs assessment, where we brainstormed aspirational research studies and identified three core infrastructure requirements: coordinated data collection and measurement systems, sustainable large-scale research-practice partnership frameworks, and knowledge repositories combined with professional learning networks. In Phase 2 we designed an integrated solution to address these needs. The INTERACT Incubator’s solution differs from existing research infrastructure because ours was systematically designed to address research needs identified by the field itself, rather than building research services on top of existing research capacities or operational systems. This commentary documents a consensus vision for novel infrastructure that would enable the research needed to achieve meaningful progress in STEM education.

What is wrong with the peer review system? Is peer review sustainable? Useful? What other models exist? These are central yet contentious questions in today’s academic discourse. This perspective critically discusses alternative models and revisions to the peer review system. The authors highlight possible changes to the peer review system, with the goal of fostering further dialog among the main stakeholders, including producers and consumers of scientific research. Neither our list of identified issues with the peer review system nor our discussed resolutions are complete. A point of agreement is that fair assessment and efficient change would require more comprehensive and rigorous data on the various aspects of the peer review system.

Funding of curiosity-driven science is the lifeblood of scientific and technological innovation. Various models of funding allocation became institutionalized in the 20th century, shaping the present landscape of research funding. There are numerous reasons for scientists to be dissatisfied with current funding schemes, including the imbalance between funding for curiosity-driven and mission-directed research, regional and country disparities, path-dependency of who gets funded, gender and race disparities, low inter-reviewer reliability, and the trade-off between the effort and time spent on writing or reviewing proposals and doing research. We discuss possible alternative models for dealing with these issues. These alternatives include incremental changes such as placing more weight on the proposals or on the investigators and representative composition of panel members, along with deeper reforms such as distributed or concentrated funding and partial lotteries in response to low inter-reviewer reliability. We also consider radical alternatives to current funding schemes: the removal of political governance and the introduction of international competitive applications to a World Research Council alongside national funding sources. There is likely no single best way to fund curiosity-driven research; we examine arguments for and against the possibility of systematically evaluating alternative models empirically.

In the present exploratory study we investigate whether cognitive flexibility is a unitary mechanism underlying flexible behaviours across many domains or a domain-specific capacity. The literature on cognitive flexibility is divided into several research lines that do not converge. The most prominent one considers flexibility an executive function that represents the ability to switch among rules or tasks. In other research traditions it is associated with distinct components, such as the capacity to place an item into many categories (in creativity tests) or a characteristic of different cognitive or perceptual processes (e.g., flexible language use, flexibility in mathematics, perceptual flexibility). To determine whether flexibility in different domains relies on a general shared mechanism, 221 subjects from two countries (The United States and Romania, mean age 19.52 years) were tested online with several measurements from four different domains of investigation: language, mathematics, perception, and executive functions (specifically, set shifting). All tasks required some form of cognitive flexibility. In addition, we measured math anxiety to see how this relates to mathematical flexibility. The results show very few and small significant partial correlations among the tasks. They also highlight that there is no unitary overarching “executive” factor. The most prominent common factor was speed of processing for mathematical and language response times. Shifting does not seem to be a mechanism that underlies flexibility in all the investigated domains. While we acknowledge the need for replication of this study, the data suggest that the construct of shifting does not exhaust the notion of flexibility as it arises across cognitive domains.

Interactive computer simulations are commonly used as pedagogical tools to support students’ statistical reasoning. This paper examines whether and how these simulations enable their intended effects. We begin by contrasting two theoretical frameworks—dual processes and grounded cognition—in the context of people’s conceptions about statistical sampling, setting the stage for the potential benefits of simulations in learning such conceptions. Then, we continue with reviewing the educational literature on statistical sampling simulations. Our review tentatively suggests benefits of the simulations for building statistical habits of mind. However, challenges seem to persist when more specific concepts and skills are investigated. With and without simulations, students have difficulty forming an aggregate view of data, interpreting sampling distributions, showing a process-based understanding of the law of large numbers, making statistical inferences, and context-independent reasoning. We propose that grounded cognition offers a framework for understanding these findings, highlighting the bidirectional relationship between perception and conception, perceptual design features, and guided perceptual routines for supporting students’ meaning making from simulations. Finally, we propose testable instructional strategies for using simulations in statistics education.

Perceptual learning is commonly understood as conferring some benefit to the learner, such as allowing for the extraction of more information from the environment. However, perceptual learning can be biased in several different ways, some of which do not appear to provide such a benefit. Here we outline a systematic framework for thinking about bias in perceptual learning and discuss how several cases fit into this framework. We argue these biases are compatible with an understanding in which perceptual learning is beneficial, but that its benefits are tied to both a person's narrow interests and the training environment or domain, and so if there are changes to either of these, then benefits can turn into liabilities, though these are often temporary.This article is categorized under: Psychology > Learning Philosophy > Value Linguistics > Language Acquisition

Many life-influencing social networks are characterized by considerable informational isolation. People within a community are far more likely to share beliefs than people who are part of different communities. The spread of useful information across communities is impeded by echo chambers (far greater connectivity within than between communities) and filter bubbles (more influence of beliefs by connected neighbors within than between communities). We apply the tools of network analysis to organize our understanding of the spread of beliefs across modularized communities and to predict the effect of individual and group parameters on the dynamics and distribution of beliefs. In our Spread of Beliefs in Modularized Communities (SBMC) framework, a stochastic block model generates social networks with variable degrees of modularity, beliefs have different observable utilities, individuals change their beliefs on the basis of summed or average evidence (or intermediate decision rules), and parameterized stochasticity introduces randomness into decisions. SBMC simulations show surprising patterns; for example, increasing out-group connectivity does not always improve group performance, adding randomness to decisions can promote performance, and decision rules that sum rather than average evidence can improve group performance, as measured by the average utility of beliefs that the agents adopt. Overall, the results suggest that intermediate degrees of belief exploration are beneficial for the spread of useful beliefs in a community, and so parameters that pull in opposite directions on an explore–exploit continuum are usefully paired.