Jennifer E. Corbett

Yuqi Hu, Vasha DuTell, Ahna R. Girshick et al.

The classic duck-rabbit illusion reveals that when visual evidence is ambiguous, the human brain must decide what it sees. But where exactly do human observers draw the line between ''duck'' and ''rabbit'', and do machine classifiers draw it in the same place? We use semantically ambiguous images as interpretability probes to expose how vision models represent the boundaries between concepts. We present a psychophysically-informed framework that interpolates between concepts in the CLIP embedding space to generate continuous spectra of ambiguous images, allowing us to precisely measure where and how humans and machine classifiers place their semantic boundaries. Using this framework, we show that machine classifiers are more biased towards seeing ''rabbit'', whereas humans are more aligned with the CLIP embedding used for synthesis, and the guidance scale seems to affect human sensitivity more strongly than machine classifiers. Our framework demonstrates how controlled ambiguity can serve as a diagnostic tool to bridge the gap between human psychophysical analysis, image classification, and generative image models, offering insight into human-model alignment, robustness, model interpretability, and image synthesis methods.

Jaap Munneke, Jennifer E. Corbett

Synthesizing from Corbett and Munneke (2025), who demonstrated that questions originating in human-computer interaction (HCI) and game design can be answered through the theoretical toolkit of cognitive science, this perspective argues that commercial videogames represent a largely underutilized research environment at the intersection of these two fields. Cognitive science has long relied on carefully controlled laboratory paradigms to study perception, attention, and executive functioning, raising persistent questions about ecological validity. HCI, by contrast, has spent decades developing methods for studying behavior in rich, complex, interactive environments, but has been less concerned with what that behavior reveals about underlying cognitive mechanisms. Commercial videogames sit precisely at this intersection. They are cognitively demanding by design, motivating by nature, and consistent enough across players to support systematic behavioral comparison. The affordance structure of a game does the work that experimental manipulations typically require of the researcher, instantiating cognitive demands that are genuine, sustained, and meaningful to the player. We argue that perception, attention, and executive functioning can be meaningfully studied within commercial games using a minimal observational toolkit of screen recording, eye tracking, and behavioral timing. We propose an affordance-cognition mapping framework as a systematic basis for game selection and research design and offer practical methodological recommendations for researchers wishing to work in this space.