Comparing theory-driven and data-driven attractiveness models using images of real women’s faces by Holzleitner et al. (in press) JEP:HPP.

[ajstewartlang]

Paper title:

Comparing theory-driven and data-driven attractiveness models using images of real women’s faces by Holzleitner et al. (in press) JEP:HPP.

Description:

This paper contrasts the top down (i.e., theory driven) with bottom up (i.e., data driven) models of attractiveness judgements (provided by an independent group of raters). The top down approach used a number of components (e.g., facial asymmetry) that were objectively designed and measured (and were chosen due to their prevalence in theories of facial attractiveness). The data driven approach was entirely bottom-up and theory-independent. For both the group of theory-driven models and the data-driven model, rated attractiveness was the outcome measure with (for the theory-driven models) subsets of the objective measures (e.g., facial asymmetry) as predictors. Goodness-of-fit of each model was measured using root-mean-square error (RMSE) with (overall) the data-driven approach explaining more variability in the data (relative to the top-down only model). Additionally, the data-driven model had the lowest AIC indicating that even with additional predictors, the explanatory performance of the model was sufficiently improved to justify the inclusion of the full range of predictors.

Participants

• Andrew Stewart [@ajstewart_lang/github.ajstewartlang] (<https://twitter.com/ajstewart_lang https://github.com/ajstewartlang>)

Resources

Paper URL: https://psyarxiv.com/vhc5k

Data URL: https://osf.io/jurcq/

Code URL: https://osf.io/jurcq/

• reproduction/replication URL: