Emily M. Bender
Last reviewed
May 31, 2026
Sources
16 citations
Review status
Source-backed
Revision
v1 ยท 2,319 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
May 31, 2026
Sources
16 citations
Review status
Source-backed
Revision
v1 ยท 2,319 words
Add missing citations, update stale details, or suggest a clearer explanation.
Emily M. Bender is an American linguist and a professor in the Department of Linguistics at the University of Washington, where she directs the Computational Linguistics Laboratory. She works in computational linguistics and natural language processing, with research that spans multilingual grammar engineering, the documentation of datasets, and the social effects of language technology. Outside the academic literature she is widely known as a critic of the marketing and public framing of large language model systems. She popularized the phrase "stochastic parrots" through a 2021 paper she co-wrote with Timnit Gebru and others, gave her name to the "Bender Rule" on naming the languages a study covers, and co-wrote the 2025 book "The AI Con" with the sociologist Alex Hanna. [1][2][3]
Bender was born in 1973. She earned an A.B. in linguistics from the University of California, Berkeley in 1995, where she received the University Medal and was elected to Phi Beta Kappa. She completed an M.A. in linguistics at Stanford University in 1997 and a Ph.D. in linguistics at Stanford in 2000. Her dissertation, "Syntactic Variation and Linguistic Competence: The Case of AAVE Copula Absence," studied grammatical variation in African American Vernacular English, and her committee included Thomas Wasow, Penelope Eckert, John Rickford, Ivan Sag, and Arnold Zwicky. [1][4]
During 2000 and 2001 she was a lecturer and postdoctoral researcher at UC Berkeley, and from 2001 to 2002 she worked in industry at YY Technologies as a grammar engineer for Japanese. She held visiting and acting positions at Stanford and at the university's Center for the Study of Language and Information before moving to the University of Washington. [4]
Bender joined the University of Washington Department of Linguistics in 2004. She was an assistant professor from 2004 to 2010, an associate professor from 2010 to 2014, and a full professor from 2014. She has directed the Computational Linguistics Laboratory since 2004 and has served as faculty director of the professional master's program in computational linguistics since 2005. She also holds adjunct appointments in the Paul G. Allen School of Computer Science and Engineering and in the Information School, and she is a member of the university's Tech Policy Lab and Value Sensitive Design Lab. [4][5]
She held the Howard and Frances Nostrand Endowed Professorship from 2019 to 2022 and was named the Thomas L. and Margo G. Wyckoff Endowed Professor for the 2024 to 2027 period. [4]
Much of Bender's early research is in grammar engineering, the construction of formal, machine-readable descriptions of how a language's sentences are built. She works within Head-driven Phrase Structure Grammar, a constraint-based theory of syntax, and within the DELPH-IN consortium, an international group that builds shared tools and precision grammars for deep linguistic processing. [5][6]
Her best known contribution in this area is the LinGO Grammar Matrix, a project she helped start in the early 2000s. The Grammar Matrix is an open-source starter kit for building broad-coverage precision grammars that stay consistent across very different languages. It pairs a shared core of linguistic analyses with a customization system, so a linguist can answer a structured questionnaire about a language and generate the skeleton of a working grammar from it. The approach is aimed in part at less widely studied and endangered languages, and Bender has used it in teaching and in field-oriented research. Her work in this period covered the structures of languages including Japanese, Mandarin, Wambaya, and others, and she wrote two reference volumes, "Linguistic Fundamentals for Natural Language Processing" (2013) and a second volume on semantics and pragmatics with Alex Lascarides (2019). [4][5][6]
Bender is associated with a short piece of research advice now called the Bender Rule: always name the language or languages you are working on. The point is that NLP papers often study English while describing the work in general terms, which hides the chance that a method is specific to English rather than universal. By naming the language, a paper makes its scope explicit and treats English as one language among the world's several thousand rather than a default. The label spread on social media and was later written up under that name in The Gradient. [7][8]
This concern with making research scope explicit connects to her work on data documentation. In a 2018 paper in the Transactions of the Association for Computational Linguistics, "Data Statements for Natural Language Processing: Toward Mitigating System Bias and Enabling Better Science," Bender and Batya Friedman proposed that datasets used to build language technology should ship with a structured description of who produced the data, in what setting, and for what purpose. They argued that such statements would let researchers reason about how far results generalize, would surface possible sources of bias, and would improve the science of the field. The proposal became part of a wider movement toward dataset and model documentation in machine learning. [9]
In 2020 Bender and the computational linguist Alexander Koller published "Climbing towards NLU: On Meaning, Form, and Understanding in the Age of Data" at the annual meeting of the Association for Computational Linguistics. The paper makes a conceptual argument about what it would take for a machine to achieve natural language understanding. Bender and Koller draw a sharp line between form, the observable marks and sounds of language, and meaning, the relation between those forms and things outside language such as speakers' intentions and the world. Their claim is that a system trained only on form, however much of it, has no way in principle to learn meaning, because meaning is not present in the form alone. [10]
To make the point vivid the authors offer a thought experiment that became known as the octopus test. A hyper-intelligent deep-sea octopus taps an undersea cable and listens to two people on separate islands sending text messages back and forth. Over time the octopus gets very good at the statistics of the exchange and can insert plausible replies that the human correspondents do not notice. But when one islander faces a real situation, such as building a tool to fend off a bear, the octopus has no way to give useful help, because it has only ever seen the form of the messages and never connected those forms to bears, coconuts, or ropes. Bender and Koller use the story to argue that fluent output is not evidence of understanding, and to caution that rapid gains on benchmarks may be climbing the wrong hill if the goal is genuine comprehension. [10]
In March 2021 Bender presented "On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?" at the ACM Conference on Fairness, Accountability, and Transparency. Her co-authors were Timnit Gebru, Angelina McMillan-Major, and a researcher listed under the pseudonym Shmargaret Shmitchell, widely reported to be the Google researcher Margaret Mitchell. The paper reviews the risks of building ever larger language models trained on text scraped from the web. [2][11]
The paper groups its concerns into several areas. It points to the financial and environmental costs of training very large models, including energy use and carbon emissions that fall hardest on communities least likely to benefit. It argues that web-scale training data encodes the biases and exclusions of who gets to post online, so models can reproduce racist, sexist, and other harmful associations in ways that are hard to audit. It revisits the form versus meaning theme, describing a language model as a "stochastic parrot," a system that stitches together sequences of linguistic form according to probability without any reference to meaning. And it warns that fluent but ungrounded text can be used to generate misinformation at scale. The phrase "stochastic parrot" later spread well beyond the paper and was named the American Dialect Society's AI-related word of the year for 2023. [2][11][12]
The paper drew attention as much for the events around it as for its content. According to widely reported accounts, Google managers told Gebru to retract the paper or remove the names of Google-affiliated authors. Gebru did not comply, and her employment ended in December 2020 in a dispute the two sides described differently, with Gebru saying she was fired and Google saying it accepted a resignation. Mitchell left Google in early 2021. The episode became a focal point in debates about academic freedom and corporate influence over research, and it helped make the paper one of the most discussed documents in recent work on AI ethics. [2][11]
Bender has become one of the more visible academic skeptics of the language used to sell generative AI. She argues that the term artificial intelligence is often a marketing label that lends a sense of agency and understanding to what are, in her account, systems for predicting likely sequences of words. She is critical of claims that current systems are approaching artificial general intelligence, and she has urged that the people building and selling these systems, rather than the systems themselves, be held responsible for their effects. She has also questioned applications in areas such as search, health, and education where confident but unreliable text can cause harm. These are her stated positions, and they are contested by researchers who hold that large models display forms of understanding or reasoning. [3][13][14]
Since 2022 Bender has co-hosted the podcast "Mystery AI Hype Theater 3000" with the sociologist Alex Hanna of the Distributed AI Research Institute. The show examines specific claims and products and tries to separate documented capabilities from promotional language. In May 2025 Bender and Hanna published a book that grew out of the podcast, "The AI Con: How to Fight Big Tech's Hype and Create the Future We Want," with Harper in the United States and The Bodley Head in the United Kingdom. The book argues that much of what is sold as AI repackages ordinary automation while concentrating power in a few large firms, and it offers readers ways to question such claims. Reviews were mixed, with some praising its contrarian stance and others finding its tone polemical or its attention to possible benefits limited. [3][14][15]
Bender was elected a Fellow of the American Association for the Advancement of Science in 2022. She was included in the 2021 list of 100 Brilliant Women in AI Ethics and in the inaugural 2023 TIME100 AI list of the most influential people in artificial intelligence. Within her field she was elected to the rotating leadership of the Association for Computational Linguistics, serving as president in 2024 and delivering the presidential address at the society's meeting in Bangkok in August 2024. Earlier honors include a National Science Foundation Graduate Research Fellowship and the University Medal from UC Berkeley. [1][4][16]
| Field | Detail |
|---|---|
| Full name | Emily Menon Bender |
| Born | 1973 |
| Nationality | American |
| Fields | Linguistics, computational linguistics, natural language processing |
| Education | UC Berkeley (A.B. linguistics, 1995); Stanford University (M.A. 1997; Ph.D. 2000) |
| Doctoral advisors | Thomas Wasow and Penelope Eckert |
| Institution | University of Washington (since 2004) |
| Positions | Professor of Linguistics; Director, Computational Linguistics Laboratory; Faculty Director, M.S. in Computational Linguistics |
| Endowed chair | Thomas L. and Margo G. Wyckoff Endowed Professor (2024 to 2027) |
| Known for | LinGO Grammar Matrix; the Bender Rule; data statements; "Climbing towards NLU"; "On the Dangers of Stochastic Parrots"; popularizing "stochastic parrots" |
| Books | "Linguistic Fundamentals for Natural Language Processing" (2013, 2019); "The AI Con" (2025, with Alex Hanna) |
| Podcast | "Mystery AI Hype Theater 3000" (co-host, since 2022) |
| Honors | AAAS Fellow (2022); TIME100 AI (2023); ACL President (2024) |