Research

This book chapter provides a gentle introduction to four paradigms for quantifying individual differences in network representations: the error paradigm, the free-recall paradigm, the structural learning paradigm, and the statistical learning paradigm.

Social evaluation, such as peer recognition, is integral to social and professional life, and being nominated for leadership roles or professional awards can significantly advance one’s career. Yet peer recognition often disadvantages those who would benefit most. To address these disparities, many organizations have introduced diversity guidelines into nomination processes, most commonly by diversifying the pool of nominators. The underlying assumption is that demographic diversity among evaluators will counteract in-group favoritism and yield a more diverse pool of nominees. Empirical support for this practice, however, has been mixed. We theorize that nomination bias arises both from biased evaluation of the considered candidates (evaluation bias) and from who is considered in the first place (recall bias). Through survey experiments and cognitive modeling, we find that evaluation bias accounts for only a small share of nomination bias; a much larger share arises from recall bias. Further analysis suggests that recall bias is rooted in network proximity rather than socio-demographic similarity, and is immune to changes in nomination tasks. These findings challenge prevailing diversity guidelines on nomination procedures. They suggest that effective efforts should focus less on diversifying nominators per se and more on diversifying their networks and implementing interventions that encourage nominators to consider beyond their close networks.

Researchers have long debated how accurately individuals perceive their own social networks. While sociologists often view individuals as unreliable informants of their social ties, neuroscientists find that the brain reliably encodes network structures. I resolve this “informant accuracy paradox” by distinguishing between dyadic accuracy (individual ties) and structural accuracy (global patterns). Drawing on contemporary theories of the brain as a predictive machine, I argue that human memory functions as a compression device that is adaptive for social navigation and inference. Using simulations, I demonstrate that while recall is lossy, frequently forgetting existing ties (omission errors) or falsely inferring non-existing ties through triadic closure (commission errors), these commission errors actually preserve and sometimes even improve the accuracy of inferred network centrality (i.e., who is more central in the network). These results reconcile pessimistic sociological assessments of informant inaccuracy with optimistic neuroscientific evidence on neural encoding of social networks. More broadly, they suggest that apparent inaccuracy in dyadic recall reflects an efficient cognitive strategy: compressing complex social information while preserving what matters for social navigation.

When identifying market opportunities for novel technologies, ventures must balance search breadth (novelty/diversity) and depth (usefulness). Yet established human-led approaches—technology broadcasting and expert search—typically prioritize one over the other. We develop and validate an autonomous generative artificial intelligence (GenAI) agent and assess its ability to address this tension. We find that GenAI-generated opportunities score higher on novelty and usefulness and exhibit greater consistency, suggesting the exploration of potential underlying mechanisms. We extend entrepreneurship research and specifically opportunity identification by introducing autonomous GenAI-based search as a distinct type integrating breadth and depth, challenging the assumed trade-off in opportunity search.

Selective attention is commonly regarded as an unfortunate inconvenience of humans’ bounded rationality. As Simon (1971) put it, “a wealth of information creates a poverty of attention” (p. 41). A long tradition in organization theory thus centers on how to accommodate limited attentional capacity, by channeling, structuring, and sequentially allocating attention to issues and answers. In this paper, we challenge this view and suggest that selective attention can be a desirable feature, and therefore, a choice, for organizational learning. We show that selective attention facilitates learning by balancing approximation errors and estimation errors, and under rivalrous access to learning opportunities (e.g., competing for experience to learn from), can generate sustained competitive advantage.