Person, Human, Neither: The Dehumanization Potential of Automated Image Tagging
Open accessAbstract
Following the literature on dehumanization via technology, we audit six proprietary image tagging algorithms (ITAs) for their potential to perpetuate dehumanization. We examine the ITAs' outputs on a controlled dataset of images depicting a diverse group of people for tags that indicate the presence of a human in the image. Through an analysis of the (mis)use of these tags, we find that there are some individuals whose 'humanness' is not recognized by an ITA, and that these individuals are often from marginalized social groups. Finally, we compare these findings with the use of the 'face' tag, which can be used for surveillance, revealing that people's faces are often recognized by an ITA even when their 'humanness' is not. Overall, we highlight the subtle ways in which ITAs may inflict widespread, disparate harm, and emphasize the importance of considering the social context of the resulting application.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
Image captioning is prone to encode and reproduce racism and discrimination.
“In computer sciences, it is well known that image captioning is prone to encode and reproduce racism and discrimination.”
From ‘We can see a savage’: a case study of the colonial gaze in generative AI algorithms (2026) · cites this paper as Barlas et al. 2021
Cites
Barocas, S., & Selbst, A. D. (2014) · doi:10.2139/ssrn.2477899 · Cited by 3,048 · see citing works
Algorithm appreciation: People prefer algorithmic to human judgment
Logg, J. M., Minson, J. A., & Moore, D. A. (2019) · Organizational Behavior and Human Decision Processes · doi:10.1016/j.obhdp.2018.12.005 · Cited by 2,061 · see citing works
Lee, M. K. (2018) · Big Data & Society · doi:10.1177/2053951718756684 · Cited by 1,338 · see citing works
Wang, Y., & Kosinski, M. (2018) · Journal of Personality and Social Psychology · doi:10.1037/pspa0000098 · Cited by 766 · cited 2× · see citing works
“Other types of computer vision applications have emerged that analyze people’s images (often their face) to infer an abstract quality about them – like personality [53], gender [23], sexual orientation [50], or emotional state [5] – even if it has not been tested whether a connection between the abstract quality and t…”
BackgroundThe Misgendering Machines: Trans/HCI Implications of Automatic Gender Recognition
Keyes, O. (2018) · Proceedings of the ACM on Human-Computer Interaction · doi:10.1145/3274357 · Cited by 505 · see citing works
Scheuerman, M. K., Paul, J. M., & Brubaker, J. R. (2019) · Proceedings of the ACM on Human-Computer Interaction · doi:10.1145/3359246 · Cited by 252 · see citing works
Algorithmic Authority: The Case of Bitcoin
Lustig, C., & Nardi, B. (2015) · 2015 48th Hawaii International Conference on System Sciences · doi:10.1109/hicss.2015.95 · Cited by 166 · see citing works
The Profiling Potential of Computer Vision and the Challenge of Computational Empiricism
Goldenfein, J. (2019) · Proceedings of the Conference on Fairness, Accountability, and Transparency · doi:10.1145/3287560.3287568 · Cited by 21 · cited 6× · see citing works
“…argues that the “previous claims about the ‘mechanical objectivity’ of the camera have been displaced into the computational objectivity of statistics” through computer vision applications using ML, where neutrality can be used (consciously or not) to “launder” ideologies that may be harmful [17].”
BackgroundInfluential
Flagged first, then by citations
Cited by
Sorted by citations
‘We can see a savage’: a case study of the colonial gaze in generative AI algorithms
Alenichev, A., Shaffer, J. D., Kingori, P., Grietens, K. P., Muldoon, J., & Rocher, L. (2026) · AI & SOCIETY · doi:10.1007/s00146-025-02685-0 · Cited by 3 · see citing works
“…also process complex relationships between human subjects and objects with their own social context and background (Sarhan and Hegelich 2020), which creates openings for the reproduction of dominant knowledge formations, bias, and representational harms in describing images (Barlas et al. 2021).”
- Authors
- Barlas, P., Kyriakou, K., Kleanthous, S., & Otterbacher, J.
- Affiliations
- CYENS Centre of Excellence, Nicosia, Cyprus
- Open University of Cyprus (CY)
- Venue
- Proceedings of the 2021 AAAI/ACM Conference on AI, Ethics, and Society
- Publisher
- ACM
- Published
- 2021-07-21
- Pages
- 357-367
- ISSN
- 3065-8365 (ISSN-L)
- OpenAlex
- W3192405527
- Cited by
- 6
- References
- 8
- Citations by year
- 2025: 3 · 2023: 3
- FWCI
- 0.68
- Type
- conference paper
- Language
- en
- Open access
- green
- zenodo.org
- Locations
- Proceedings of the AAAI/ACM Conference on AI Ethics and Society (publishedVersion)
- Zenodo (CERN European Organization for Nuclear Research) (submittedVersion, cc-by)
- Kypseli (submittedVersion)
- Funders
- Horizon 2020 Framework Programme (739578, 810105)
- the Government of the Republic of Cyprus through the Deputy Ministry of Research, Innovation and Digital Policy
- Topics
- Social Sciences › Safety Research › Ethics and Social Impacts of AI
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Face recognition and analysis
- Social Sciences › Business, Management and Accounting › Organizational Behavior and Human Resource Management › AI and HR Technologies
- Keywords
- dehumanization, human detection, marginalized communities, face recognition, surveillance, algorithmic bias, social context of technology, human/non-human distinction
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
