Beyond Text: Using AI-Generated Visual Conjoints to Study Gender and Housework Attribution

Despite substantial gender convergence in education and employment, women continue to perform a disproportionate share of housework. We employ a novel visual conjoint experiment to isolate the normative mechanisms underlying this persistent inequality. Using AI-generated photorealistic images, we systematically vary the tidiness of domestic spaces, room type, source of mess, socioeconomic status, and the gender and race/ethnicity of occupants, alongside text describing couples’ employment arrangements. A quota sample of 2,994 U.S. respondents each evaluated five vignettes, yielding 14,970 observations. We find that gender effects operate primarily through responsibility attribution rather than through differential perception of messiness or anticipated social judgment. Women are assigned significantly more cleaning responsibility than men, with the gender penalty concentrated among dual-earner couples. Child-caused mess is perceived as messier than adult-caused mess yet carries reduced social consequences, suggesting that it operates as a legitimating excuse. Our findings suggest that gender equality in paid work is necessary for achieving gender equality in housework, but that it is not sufficient, and that this gap will persist absent changes in normative expectations around responsibility for housework.