Abeba Birhane is an Ethiopian-born cognitive and computer scientist based in Ireland whose work spans complex adaptive systems, machine learning, artificial intelligence ethics, algorithmic bias and accountability, and critical race studies.^[22] She completed a BSc and BA at Bahir Dar University in Ethiopia, followed by an MSc and PhD at University College Dublin. Her research analyzes the societal and ethical impacts of artificial intelligence, examining the risks of large datasets, the biases of computational models, the mechanisms reproducing systemic bias, racism, and misogyny in emerging technologies, and the limitations of applying automated methods to human and social phenomena.^[41]

Birhane gained international attention for her critical audits of large-scale image and multimodal datasets commonly used to train AI systems; her collaborative work with Vinay Prabhu revealed that prominent datasets, including ImageNet, 80 Million Tiny Images, and datasets used in Stable Diffusion image generation, contained racist and misogynistic labels, harmful stereotypes, and offensive or degrading images.^{[4,5,18,19,20,22]} These findings directly contributed to the shutdown of the MIT dataset 80 Million Tiny Images.^[4,5]

In 2024, she founded and became the head of the AI Accountability Lab (AIAL) at Trinity College Dublin, alongside affiliations with University College Dublin and DeepMind. In 2023, Birhane was appointed to the United Nations High-level Advisory Body on AI.^[7,8] She has also been recognized by VentureBeat as a top innovator in computer vision and was named in 2023 to the inaugural TIME100 AI list by TIME magazine as one of the 100 most influential people in artificial intelligence.^{[7,8,21,18,51,53]}

Abeba Birhane was born in Ethiopia, where she grew up in Bahir Dar, and later settled in Ireland.^[3,22,38,23] Her academic background combines philosophy, cognitive science, and computer science; she first studied psychology, earning a Bachelor of Science, and then philosophy, earning a Bachelor of Arts, from The Open University.^[3,23] In 2015, she completed her Master of Science in cognitive science at University College Dublin.^{[24,25,32,22,23,39,54,55]} She completed her Ph.D. in cognitive science at the Complex Software Lab in the School of Computer Science at University College Dublin in 2021 according to most sources, or in 2022 according to the Italian edition, focusing her research on the relationship between complex systems, human behavior, computational methods, and audits of training datasets.^{[9,24,25,32,22,23,39,54,55]}

An important part of her work questions the idea that the data used to train artificial intelligence systems are neutral representations of reality, investigating how decisions regarding data collection, classification, and reuse can incorporate social inequalities and later reproduce them in models.^[33] She discovered that AI algorithms tend to disproportionately affect vulnerable groups such as the elderly, immigrants, and children, and her research on relational ethics won the best paper award at the NeurIPS Black in AI workshop in 2019.^[20] She has also studied and written about algorithmic colonization driven by corporate agendas, and her work on decolonizing computational sciences addressed inherited oppressions in current systems, especially toward non-white women.^[19,20]

In 2020, Birhane and Vinay Prabhu, principal machine learning scientists at UnifyID, published a paper examining the problematic collection, labeling, and classification of data in large image datasets, revealing racist and misogynistic labels, insults, and offensive images, which led MIT to voluntarily and formally withdraw the 80 Million Tiny Images dataset.^[18,19] She also participated in audits of large datasets used in AI research, documenting the presence of hate content, stereotypes, and other problematic materials in LAION-400M and LAION-2B-en, and contributed to the study of Twitter's automated image cropping system to investigate potential differences in the treatment of faces and people.^[1] Birhane collaborated with Rediet Abebe, George Obaido, and Sekou Remy on research published at the ACM Conference on Fairness, Accountability, and Transparency, which identified significant power imbalances in data sharing processes in Africa. She served as a Senior Fellow at the Mozilla Foundation in the Trustworthy AI program between 2022 and 2023, continuing research on AI system evaluation, algorithmic accountability, and the theoretical foundations of artificial intelligence.^[2,10,6] She works as a researcher at the ADAPT Research Centre and the School of Computer Science and Statistics at Trinity College Dublin, where in 2024 she founded the AI Accountability Lab research group focusing on audits of AI models and training datasets.^[10,6,33]

Birhane's research focuses on the accountability of artificial intelligence systems, integrating cognitive science, critical race studies, and decoloniality theory to examine how emerging AI technologies shape individuals, society, and local communities.^[18] Her investigations revealed that AI algorithms tend to disproportionately and severely impact vulnerable and disempowered groups, including older workers, transgender people, immigrants, and children. Birhane has also shaped the debate on algorithmic colonization driven by corporate agendas, examining how technological systems reproduce colonial and historical power logics where corporate goals prevail over marginalized communities, thereby embedding systemic oppressions that disproportionately affect women of color and Black women.^{[13,27,41,57]} As an example of these imbalances, she collaborated with Rediet Abebe, George Obaido, and Sekou Remy on research presented at the ACM Conference on Fairness, Accountability, and Transparency (FAccT) regarding barriers to data sharing in Africa, finding significant power imbalances where African countries risk losing control of their data to external actors without proportional benefits and face systematic underrepresentation even when baseline data originates from Africa.^{[32,14,6,41,62]}

A major focus of Birhane's empirical research is the auditing of large-scale image and multimodal datasets commonly used to train computer vision algorithms and AI systems.^[31] In 2020, Birhane and Vinay Uday Prabhu, chief machine learning scientist at UnifyID, published a study examining the problematic data collection pipelines, labeling, classification, and societal consequences of massive image datasets such as ImageNet and MIT's 80 Million Tiny Images, which had been used to develop thousands of AI systems.^{[4,5,28,29,18,31,59]} They discovered that these datasets contained numerous racist and misogynistic labels, slurs, and offensive, pornographic, or malicious images, which led MIT to publicly apologize and voluntarily and formally take down the 80 Million Tiny Images dataset, raising awareness within the AI community regarding data hygiene and ethics.^{[11,3,4,5,29,30,31,18]} In 2021, they conducted research on Google's JFT-300M dataset, raising ethical concerns over the presence of non-consensual images resulting from questionable collection practices.^[12] During her Mozilla fellowship in 2022–2023 and in later audits of large datasets such as LAION-400M and LAION-2B-en, Birhane demonstrated that as web datasets scale, the proportion of hateful and offensive content and resulting negative stereotyping in multimodal generative models also increases.^[12,47] She also investigated automated image-cropping algorithms used by platforms including Twitter, Apple, and Google, identifying systematic preferences for white subjects and a tendency to objectify women by emphasizing physical features rather than faces.^[41]

On a theoretical level, Birhane developed a critique of the value assumptions embedded in machine learning; in a distinguished paper presented at the 2022 ACM FAccT conference analyzing one hundred highly cited papers from ICML and NeurIPS, she and her co-authors showed that dominant field values such as performance, efficiency, and generalization concentrate power in already advantaged institutions while neglecting social consequences.^[41] In 2024, she founded and was appointed to lead the AI Accountability Lab (AIAL) at Trinity College Dublin, where researchers and civil society partners develop auditing procedures for AI systems and training datasets to support evidence-based policy regulation.^{[6,14,33,45,46]} In October 2023, United Nations Secretary-General António Guterres appointed Birhane as one of 32 experts on the newly established UN AI Advisory Body tasked with advising on global AI governance, and she is also a member of the Irish AI Advisory Council.^[48,49] Her research on relational ethics earned the Best Paper Award at the NeurIPS Black in AI workshop in 2019.^{[15,26,41,56]} Her other honors include being named to the 100 Brilliant Women in AI Ethics in 2021, receiving the Lero Director’s Prize in 2022, being recognized by VentureBeat as a leading innovator in computer vision, being named to the TIME100 AI list of the 100 most influential people in AI in 2023, and being featured in New African Magazine's 100 Most Influential Africans in 2023/2024.^[15,17,21]

In 2019, she received the Best Paper Award at the NeurIPS Black in AI workshop for research on relational ethics.^[34,41] In 2020, she received the VentureBeat AI Innovations Award in the Computer Vision Innovation category alongside Vinay Prabhu (Vinay Uday Prabhu) for their study on image datasets.^[35,21] She was named a 100 Brilliant Women in AI Ethics Hall of Fame honoree in 2021.^[36,35]

In 2022, she received the Distinguished Paper Award at the FAccT conference for the paper The Values Encoded in Machine Learning Research, as well as the Lero Director's Prize for PhD and postdoctoral contribution.^[37,36,50] In 2023, she was included in TIME100 AI, TIME magazine's inaugural list of the 100 most influential people in the field of artificial intelligence.^[21]