CURRENT PROJECTS

How does emotion affect the beauty of images and music?

We are familiar with the phenomenon that sad music is often considered beautiful and pleasurable. Can we quantify this effect? Does it hold for images too? Does the subject’s mood play a role? Results indicate that levels of happiness and sadness expressed by images and music tend to boost beauty judgments cast on those objects – though the sadness effect is stronger for music than images – and participants' self-reported emotions interact with these relationships.

Bruns, A., Pombo, M., Ripollés, P., & Pelli, D. G. (2023). Emotions of subject and object affect beauty differently for images and music. Journal of Vision, 23(13), 6–6.

Bruns, A., Pombo, M., Ripollés, P., & Pelli, D. G. Beauty and emotion of music and images: Emotions make more emotional objects more beautiful. Talk at the 2022 International Association for Empirical Aesthetics meeting, Philadelphia, PA.

Empathy strengthens beauty

It is widely supposed that empathy affects aesthetic experience. Given this, we wondered whether we could quantify the effect of empathy on our aesthetic and emotional response to images and music. We used path modeling to show that higher-empathy participants experience more emotion and beauty, and that these relationships tend to be stronger in the music than image domain. Presented as a talk in the symposium described below.

Bruns, A., & Pelli, D. G. (2025). Empathy strengthens beauty. Preprint.

Bruns, A., & Pelli, D. G. (2025). Empathy strengthens beauty. Poster at the Society for Affective Science meeting, Portland, OR.

Bruns, A., Chamberlain, R., Drake, J. E., & Sherman, A. Emotion in art experience. Symposium at the 2024 Division 10 of the American Psychological Association (Society for the Psychology of Aesthetics, Creativity, & the Arts) meeting, Denton, TX.

Measuring correlations between visual phenomena

In addition to studying aesthetics, I'm working on a project exploring a general problem in psychological science: When researchers correlate two measures across many observers, measurement noise diminishes the observed correlation. This bias is inconsequential if merely trying to reject a no-correlation hypothesis, but it poses an obstacle to estimating the true correlation. Crowding distance (the minimum spacing required to recognize an object in clutter) is correlated with the size of V4, informational masking (“auditory crowding”), and reading speed. Using visual crowding and reading speed as an example, we show how to estimate correlations with minimal error relative to the ground truth. 

Spearman (1904) provides an unbiased estimator of correlation, a promising starting point. However, while it eliminates bias, it increases variance. We care about total error, including bias and variance, both of which depend on sample size, the true correlation, and the measures’ joint reliability. While sample size is under the experimenter’s control, the latter two factors are not. Researchers can collect preliminary data to estimate them. We provide a formula that takes estimated correlation and reliability as inputs and returns the sample size required for the unbiased estimator to have lower total error than the raw sample correlation.

Bruns, A., Hu, F., Kim, M., Winawer, J., & Pelli, D. G. (2025). Measuring correlations in vision. Poster at the 2025 Vision Science Society meeting, St. Pete Beach, FL. 

Deep network representation of art style similarity judgments

This project explores the ways humans and machines judge the visual similarity of paintings–in terms of their styles, their content, and holistically. We analyze in tandem the results from a behavioral study with human participants and the performance of a deep neural network trained to judge painting similarity. Results suggest that neural networks can approximate human style similarity scores only for pairs of paintings that match in both subject and style or that match in neither (but not those that match in one but not the other). We thank Dr. Brenden Lake and Dr. Denis Pelli for their guidance and feedback.

Bruns, A., Gao, M., Dendukuri, A., & Eubank, J. Deep Network Representation of Art Style Similarity Judgments. Poster at the 2023 Vision Science Society meeting, St. Pete Beach, FL. 

PAST PROJECTS

Predicting aesthetic evaluation of artwork using PyTorch neural networks and Scikit-Learn regression

MAY 2022

Designed model that uses Minted's consumer ratings of 7,000 art pieces to predict new artworks’ ratings with error rates 82% better than random.

Minted is an art and stationery e-commerce company that crowdsources art and design from independent artists through a challenge process. Thousands of art submissions receive >1 million ratings from their consumer base each year. Minted merchants (my former role) curate collections based on these ratings. All merchants are intimately familiar with the kinds of image features that tend to result in high ratings. 

I've designed models that show reasonable success predicting consumer ratings of new artworks. 

You can view the slides here that served as my preliminary presentation of this work.

Perspectives on aesthetics

MAY 2022

This website project, produced for a digital humanities course, surveys perspectives on aesthetics in the two disciplines that study it most directly, philosophy and psychology. It examines these disciplines' salient topics and rhetoric using topic modeling, and then it describes the work being done across these disciplines, ultimately making a case for the value of a cross-disciplinary aesthetics. 

The effect of culture and emotion on aesthetic response to visual art and music   

DECEMBER 2021

Completed pilot stages of study that aimed to understand the effect of stimulus culture of origin (Chinese or Western) and emotional valence on aesthetic response to both visual art and music. In particular, is culture of origin or valence dominant in judgments of “liking” and beauty of art and music? Does the aesthetic effect of culture of origin and valence depend on perceptual modality? I anticipated that stimulus culture of origin would be more important to pleasure and beauty responses than valence, i.e. that Western participants would prefer Western over non-Western stimuli regardless of valence. 

This pilot was designed to support a broader project that asks the following question: How does culture congruence and valence congruence within pairs of songs and visual artworks influence aesthetic response to the pairs? i.e. When showing participants a visual artwork as they listen to a song, do pleasure and beauty ratings go up when the art pieces and songs within pairings are from the same culture and/or have similar valence? Is this mediated by other factors like familiarity with the stimulus, participant enculturation, and/or art or music training? In response to these questions, I expected that participants with little/no music and art training will prefer cultural and valence congruence, but participants with training will show mixed response.

Pilot data revealed exciting valence effects, which became the focus of our current work (see the top of this page). We hope to incorporate a cross-cultural component to future work.

Embodied Viewing: Dutch Golden Age Painting, Minimalist Sculpture, and Physical Space  

DECEMBER 2020

In 2019 and 2020 I turned my Boston Art Review article below into a 15-page non-traditional art history essay, which helped gain me admittance to several top art history Master's programs.

You can view the essay here. 

Editorial in the Boston Art Review

APRIL 2018   

Boston Art Review (BAR) is an independent publication committed to facilitating active discourse around contemporary art in Boston.

You can view my publication here.

CalPerformances "Front Row" Onstage Panel of Bay Area Comedians

MARCH 2018

Co-produced the 2018 “Front Row” onstage panel of comedians, which engaged issues of Asian-American identity.

You can view the event program here and the press release here.

The Value of Simplicity in a Mathematical Explanation: Plan for an Empirical Study

DECEMBER 2016

Developed and presented a plan for a study under Ny Vasil in the Concepts and Cognition Lab that asks to what extent the level of simplicity of a math proof impacts its educational and aesthetic value. 

I shifted my research focus to art theory and perception in early 2017, and the project was picked up by others in the lab.

You can view the plan report here.