Automated Inference on Criminality Using Face Images, Xiaolin Wu and Xi Zhang's 2016 arXiv preprint, claims roughly ninety percent accuracy distinguishing criminal from non-criminal faces using a machine learning classifier. Its strength is that it documents an actual claimed experiment and its stated method in the authors' own words. Its limit is that it was never published in a peer reviewed venue and was widely criticized for undisclosed photo sourcing and unruled out confounds, criticism a reader should weigh at least as heavily as the paper's own claim.
Facts
Assessment
Reliability Tier3
Reliability tier 3: a preprint never published in a peer reviewed venue and widely criticized for undisclosed sourcing and unruled out confounds. NotesClaimed ~90% accuracy distinguishing criminal from non-criminal faces; criticized for undisclosed photo sourcing and no ruled-out confounds. Never published in a peer-reviewed venue.
Citation
AuthorXiaolin Wu, Xi Zhang
PublisherarXiv preprint
Publication Year2016
Source Typepreprint, not peer-reviewed
Claims Backed By This Source (9 claims)
This source backs 9 claims across the atlas. As facts: 7 well-attested, 1 popular myth - corrected. Plus 1 entities citing it as a general reference with no single fact or relationship attached.
Disposition By Topic
- Sources, 7 claims: 7 well-attested.
- Disciplines, 2 claims: 1 popular myth - corrected, 1 general references.
Popular myth - corrected
1Reader Challenges (0 open reader challenges)
No disputes yet. Spotted an error or a better source? Open the first one.
Sign in to dispute this or suggest a correction.
View At A Past Year
The atlas records no dated fact of its own for this entry, so there is no other year to choose.