Hi! I'm Elijah.

I study people, and the way their decisions influence data and algorithms



My name is Elijah Mayfield and my current role is Entrepreneur-in-Residence at the Carnegie Mellon University School of Computer Science. I work with Project Olympus.

I’m also:



I work with foundations, companies, and researchers to make better decisions about algorithmic technology - particularly when working with machine learning and AI.

Current and former collaborators include:


Recent Work

CSCW 2019

In an extension of my previous work on Wikipedia’s decision-making process, I show that the machine learning model we developed for predicting the outcome of debates can be inspected for details about the human processes in the dataset that it was trained on. In particular I focus on replicating several results from prior work to show the validity of the model, then digging in on how users strategically cite policies to greater or lesser effect when advocating for their stance. I’ll be presenting the work at CSCW this fall. The dataset and source code is available under a GPLv3 license.

Elijah Mayfield and Alan W Black. Analyzing Wikipedia Deletion Debates with a Group Decision-Making Forecast Model. Proceedings of the ACM on Human-Computer Interaction 3.CSCW (2019).

ACL 2019

Several of my closest colleagues and I pulled together a survey on the key ethical issues at the intersection of NLP research, educational technology, and an equity agenda, trying to lay out a survey of the state of the art in fairness research and describe several case studies that suggest how to move beyond measuring bias alone. We presented it this summer In the Building Educational Applications workshop in Florence.

Elijah Mayfield, Michael Madaio, Shrimai Prabhumoye, David Gerritsen, Brittany McLaughlin, Ezekiel Dixon-Román, and Alan W Black. Equity Beyond Bias in Language Technologies for Education. Workshop on Innovative Use of NLP for Building Educational Applications at the Association for Computational Linguistics (ACL).

NAACL 2019

At the workshop on Computational Social Science, I presented a series of new tasks on small group discussion and decision-making, using real-world data from Wikipedia. The tasks themselves are interesting, including well-known work like stance classification as well as new measures of impact of individual people or contributions. The broader goal of the work is to use what we know from behavioral science to support spaces with better, more equitable group decision-making in general, not just online.

Mayfield, E. and Black, A.W. Stance Classification, Outcome Prediction, and Impact Assessment: NLP Tasks for Studying Group Decision-Making. Workshop on Natural Language Processing + Computational Social Science at the North American Association for Computational Linguistics (NAACL).

CHI 2019

My collaborator Diyi Yang built a machine learning model for understanding how patients with cancer and their caregivers use online communities. People want different things from websites, like information about an upcoming treatment, emotional support, or a community to identify with. The model Diyi implemented takes a quantitative look at those needs, in pursuit of algorithms that understand how people grow into different roles over time. This work was awarded a Best Paper Honourable Mention.

Yang, D., Kraut, R., Smith, T., Mayfield, E., & Jurafsky, D. (2019). Seekers, Providers, Welcomers, and Storytellers: Modeling Social Roles in Online Health Communities. Proceedings of the ACM Conference on Human Factors in Computing Systems.

CSCW 2018

I joined a workshop organized by Christine Wolf, Haiyi Zhu, Julia Ballard, Min Kyung Lee, and Jed Brubaker. In a full-day, interactive design activity, the group attempted to forecast the future of algorithmic involvement in peoples’ day-to-day lives, and to propose some ways researchers could help to safeguard against negative impacts. My contribution focused on online communities, studying how policy-driven arguments on Wikipedia are used as a tool by users trying to win debates, and what we might learn from them when building algorithmic decision support tools.

Mayfield, E. & Black, A.W. Constraining Decision-Making Over Time with Categories and Policies. In the Workshop on “Participation” in Data-driven Algorithmic Ecosystems at the ACM Conference on Computer-Supported Collaborative Work.

For more, see my full list of publications.



I’m always available for research projects, especially where I can provide ML/AI expertise for social science questions, and especially with natural language data.

I also occasionally accept short-term consulting contracts.


Please complete the form below

Name *