2019 Conference on Cognitive Computational Neuroscience

13-16 September 2019    Berlin, Germany

Technical Program

Paper Detail

Paper: PS-1B.15
Session: Poster Session 1B
Location: H Fläche 1.OG
Session Time: Saturday, September 14, 16:30 - 19:30
Presentation Time:Saturday, September 14, 16:30 - 19:30
Presentation: Poster
Publication: 2019 Conference on Cognitive Computational Neuroscience, 13-16 September 2019, Berlin, Germany
Paper Title: Comparing facets of behavioral object representation: implicit perceptual similarity matches brains and models
Manuscript:  Click here to view manuscript
License: Creative Commons License
This work is licensed under a Creative Commons Attribution 3.0 Unported License.
DOI: https://doi.org/10.32470/CCN.2019.1395-0
Authors: Caterina Magri, Talia Konkle, Harvard University, United States
Abstract: The similarity space of objects has been extensively used as a tool to relate representations among minds, brains, and models. However, the psychological construct of “similarity” is not well defined – objects can be similar in different ways. Here, we explored the similarity among inanimate objects, varying the instructions and task, and compared these to deep net representations and human brain responses. Specifically, we used a typical unguided sorting task in which participants drag and drop similar items nearby; a shape-guided sorting task, in which participants are explicitly instructed to arrange objects by shape similarity; and a pairwise-visual search task, in which participants have to find one target amongst others items, measuring similarity implicitly through reaction time. Our results show that (i) there are clear differences in the measured similarity space of objects across tasks, and (ii) the implicit similarity measured by visual search was better reflected in both deep net fits across all the early layers, and more extensively along the ventral visual stream. Broadly, these results highlight that different kinds of similarity can be manifest in different behavioral tasks, highlighting a rich space for elaborating the ways in which we explore representational matches between minds, brains, and models.