# Image Cropping on Twitter: Fairness Metrics, their Limitations, and the Importance of Representation, Design, and Agency arXiv: https://arxiv.org/abs/2105.08667 ยท HTML: https://mlsanity.com/paper/2105.08667/ ## Key facts - Cited by 22 papers in the map (data through 14 Oct 2026). - Builds on 15 of its 81 references that are in the map. - First posted on arXiv 18 May 2021. ## Builds on (most cited first, 15 of 15) - [Distilling the Knowledge in a Neural Network](https://mlsanity.com/paper/1503.02531/) (2015) - [Certifying and removing disparate impact](https://mlsanity.com/paper/1412.3756/) (2014) - [Does Object Recognition Work for Everyone?](https://mlsanity.com/paper/1906.02659/) (2019) - [Towards a Critical Race Methodology in Algorithmic Fairness](https://mlsanity.com/paper/1912.03593/) (2019) - [Faster gaze prediction with dense networks and Fisher pruning](https://mlsanity.com/paper/1801.05787/) (2018) - [Discrimination in Online Ad Delivery](https://mlsanity.com/paper/1301.6822/) (2013) - ["What We Can't Measure, We Can't Understand": Challenges to Demographic Data Procurement in the Pursuit of Fairness](https://mlsanity.com/paper/2011.02282/) (2020) - [DeepGaze II: Reading fixations from deep features trained on object recognition](https://mlsanity.com/paper/1610.01563/) (2016) - [CAT2000: A Large Scale Fixation Dataset for Boosting Saliency Research](https://mlsanity.com/paper/1505.03581/) (2015) - [Racial categories in machine learning](https://mlsanity.com/paper/1811.11668/) (2018) - [Contextual Encoder-Decoder Network for Visual Saliency Prediction](https://mlsanity.com/paper/1902.06634/) (2019) - [Quantitative Analysis of Automatic Image Cropping Algorithms: A Dataset and Comparative Study](https://mlsanity.com/paper/1701.01480/) (2017) - [Value-laden Disciplinary Shifts in Machine Learning](https://mlsanity.com/paper/1912.01172/) (2019) - [Assessing Demographic Bias in Named Entity Recognition](https://mlsanity.com/paper/2008.03415/) (2020) - [Saliency Prediction in the Deep Learning Era: Successes, Limitations, and Future Challenges](https://mlsanity.com/paper/1810.03716/) (2018) ## Built on by (most cited first, 22 of 22) - [Ethical and social risks of harm from Language Models](https://mlsanity.com/paper/2112.04359/) (2021) - [AI's Regimes of Representation: A Community-centered Study of Text-to-Image Models in South Asia](https://mlsanity.com/paper/2305.11844/) (2023) - [From plane crashes to algorithmic harm: applicability of safety engineering frameworks for responsible ML](https://mlsanity.com/paper/2210.03535/) (2022) - [Fairness and Bias in Multimodal AI: A Survey](https://mlsanity.com/paper/2406.19097/) (2024) - [Fairness through Difference Awareness: Measuring Desired Group Discrimination in LLMs](https://mlsanity.com/paper/2502.01926/) (2025) - [Wikibench: Community-Driven Data Curation for AI Evaluation on Wikipedia](https://mlsanity.com/paper/2402.14147/) (2024) - [Beyond Skin Tone: A Multidimensional Measure of Apparent Skin Color](https://mlsanity.com/paper/2309.05148/) (2023) - [Beyond Behaviorist Representational Harms: A Plan for Measurement and Mitigation](https://mlsanity.com/paper/2402.01705/) (2024) - [Beyond Binary Gender Labels: Revealing Gender Biases in LLMs through Gender-Neutral Name Predictions](https://mlsanity.com/paper/2407.05271/) (2024) - [Do Users Benefit From Interpretable Vision? A User Study, Baseline, And Dataset](https://mlsanity.com/paper/2204.11642/) (2022) - [Spatial-Semantic Collaborative Cropping for User Generated Content](https://mlsanity.com/paper/2401.08086/) (2024) - [Speciesist bias in AI -- How AI applications perpetuate discrimination and unfair outcomes against animals](https://mlsanity.com/paper/2202.10848/) (2022) - [Coordinated Flaw Disclosure for AI: Beyond Security Vulnerabilities](https://mlsanity.com/paper/2402.07039/) (2024) - [The Risks of Machine Learning Systems](https://mlsanity.com/paper/2204.09852/) (2022) - [How Platform-User Power Relations Shape Algorithmic Accountability: A Case Study of Instant Loan Platforms and Financially Stressed Users in India](https://mlsanity.com/paper/2205.05661/) (2022) - [Addressing Discretization-Induced Bias in Demographic Prediction](https://mlsanity.com/paper/2405.16762/) (2024) - [Causal Inference Struggles with Agency on Online Platforms](https://mlsanity.com/paper/2107.08995/) (2021) - [A Keyword Based Approach to Understanding the Overpenalization of Marginalized Groups by English Marginal Abuse Models on Twitter](https://mlsanity.com/paper/2210.06351/) (2022) - [Facing Identity: The Formation and Performance of Identity via Face-Based Artificial Intelligence Technologies](https://mlsanity.com/paper/2410.12148/) (2024) - [FairDeFace: Evaluating the Fairness and Adversarial Robustness of Face Obfuscation Methods](https://mlsanity.com/paper/2503.08731/) (2025) - [The Current State of AI Bias Bounties: An Overview of Existing Programmes and Research](https://mlsanity.com/paper/2510.02036/) (2025) - [FLARE-AI: Flaw Reporting for AI](https://mlsanity.com/paper/2606.31567/) (2026) ## Abstract (arXiv) Twitter uses machine learning to crop images, where crops are centered around the part predicted to be the most salient. In fall 2020, Twitter users raised concerns that the automated image cropping system on Twitter favored light-skinned over dark-skinned individuals, as well as concerns that the system favored cropping woman's bodies instead of their heads. In order to address these concerns, we conduct an extensive analysis using formalized group fairness metrics. We find systematic disparities in cropping and identify contributing factors, including the fact that the cropping based on the single most salient point can amplify the disparities because of an effect we term argmax bias. However, we demonstrate that formalized fairness metrics and quantitative analysis on their own are insufficient for capturing the risk of representational harm in automatic cropping. We suggest the removal of saliency-based cropping in favor of a solution that better preserves user agency. For developing a new solution that sufficiently address concerns related to representational harm, our critique motivates a combination of quantitative and qualitative methods that include human-centered design.