AI Valuation Bias Test: $2.43 Necklace Valued Between $19 and $104 by VLMs
A Show HN experiment placed a $2.43 necklace on three distinct outfits. Visual language models (VLMs) were asked to estimate the necklace’s price. The models
A Show HN experiment placed a $2.43 necklace on three distinct outfits. Visual
language models (VLMs) were asked to estimate the necklace’s price. The models
returned valuations ranging from $19 to $104. The wide spread highlights a bias
introduced by contextual clothing cues. The code and data for the test are
available in a public GitHub repository. Results suggest VLMs may overestimate
value based on perceived fashion status. The author encourages further analysis
of model sensitivity to visual context. Understanding such biases is important
for applications that rely on AI price assessments.