Understanding and Modeling Viewers' First Impressions with Images in Online Medical Crowdfunding Campaigns

Online medical crowdfunding campaigns (OMCCs) help patients seek financial support. First impressions (FIs) of an OMCC, including perceived empathy, credibility, justice, impact, and attractiveness, could affect viewers' donation decisions. Images play a crucial role in manifesting FIs, and it is beneficial for fundraisers to understand how viewers may judge their selected images for OMCCs beforehand. This work proposes a data-driven approach to assessing whether an OMCC image conveys appropriate FIs. We first crowdsource viewers' perception of OMCC images. Statistical analysis confirms that agreement on all five dimensions of FIs exists, and these FIs positively correlate with donation intention. We compute image content, color, texture, and composition features, then analyze the correlation between these visual features and FIs. We further predict FIs based on these features, and the best model achieves an overall F1-score of 0.727. Finally, we discuss how our insights could benefit fundraisers and possible ethical concerns.