The Problem with Metrics is a Fundamental Problem for AI

By Rachel L. Thomas, D. Uminsky. 2020

48 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Consequences of Misaligned AI (2021, background): “The misalignment between what we can specify and what we want has already caused significant harms (32).”
Interactive Model Cards: A Human-Centered Approach to Model Documentation (2022, background): “, metrics [16, 67]), and we urge caution to ensure that these approaches are understandable and actionable for non-experts.”
The shaky foundations of large language models and foundation models for electronic health records (2023, methodology): “…automated metrics for evaluating the more qualitative aspects of a model ’ s “ alignment ” with human values (e.g., helpfulness or harmlessness) 86 , even as the importance of human feedback during training has been repeatedly demonstrated via techniques like Reinforcement Learning from Human…”
SECure: A Social and Environmental Certificate for AI Systems (2020, background): “This idea of a focus on single metrics (Thomas & Uminsky, 2020) to make design tradeoffs has been shown to be detrimental to the development of technology and adopting a more comprehensive approach that can internalize some of the externalities presents a great starting point to address this…”
“Everyone wants to do the model work, not the data work”: Data Cascades in High-Stakes AI (2021, methodology): “interventions highlight the importance of catching data errors using mechanisms specific to data validation, instead of using model performance as a proxy for data quality [120].”
Evaluating CLIP: Towards Characterization of Broader Capabilities and Downstream Implications (2021, background): “We need to expand our deﬁnitions of ‘better’ models to also include their possible downstream impacts, uses etc. [10], [21] We believe one step forward is community exploration to further characterize models like CLIP and holistically develop qualitative and quantitative evaluations to assess the…”