Mackenzie J. Siesel¹, Tianyu Zhang¹, Yin-ting Lin¹, Andrew B. Leber¹; ¹The Ohio State University

People choose among multiple strategies to tackle everyday visual search challenges. For example, when looking for their shiny red car in the parking lot, the most efficient strategy is to focus on its conspicuous color. However, people can also choose suboptimal strategies, like serially searching every car. Recent work has begun to reveal such suboptimal strategy usage, which may lead one to question: can strategies be improved? We have shown that explicitly informing participants of the optimal strategy boosts optimality (Zhang et al., 2024). But, how durable is the improvement? We compared optimality rates between one group of participants that was explicitly informed of the optimal strategy vs. a group that was not (N=60 per group). We found a significant decline in optimality for the instructed group over time, while there was a numerical but not significant increase over time for the non-instructed group. We then compared the linear slopes of the change in optimality over time between the two groups, revealing a significant difference. This confirmed distinct patterns of change in search strategies over time. Moreover, using a sliding window analysis, we found little evidence for rapid boosts in optimality attributable to sudden insight of the optimal strategy (c.f., see Lin & Leber, in press). Overall, explicit strategy instruction does enhance visual search performance, but this improvement appears to be short-lived. We speculate that the drive to avoid cognitive effort associated with optimal performance may eventually win out, irrespective of optimal strategy awareness. The short-lived nature of the instruction effect may explain strong test-retest reliability in optimality (Irons & Leber, 2018), as visual search strategy optimization may be largely governed by stable traits. In summary, these results shed light on how explicit knowledge and trait variables interact to drive visual search strategy use.

Acknowledgements: This work is supported by NSF BCS-2021038 to ABL.