Myra Cheng

Myra Cheng

Auto-generated profile, based on publicly available information

Myra Cheng

Stanford University, Stanford, USA

Research Focus & Metrics

Academic Career Length

5Years (2022-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
12 June 2023
Verified
Conference Paper
Open access
FBFederico Bianchi
Federico Bianchi
PKPratyusha Kalluri
Pratyusha Kalluri
EDEsin Durmus
Esin Durmus
FLFaisal Ladhak
Faisal Ladhak
MCMyra Cheng
Myra Cheng
ACAylin Caliskan
Aylin Caliskan
Machine learning models that convert user-written text descriptions into images are now widely available online and used by millions of users to generate millions of images a day. We investigate the potential for these models to amplify dangerous and complex stereotypes. We find a broad range of ordinary prompts produce stereotypes, including prompts simply mentioning traits, descriptors, occupations, or objects. For example, we find cases of prompting for basic traits or social roles resulting in images reinforcing whiteness as ideal, prompting for occupations resulting in amplification of racial and gender disparities, and prompting for objects resulting in reification of American norms. Stereotypes are present regardless of whether prompts explicitly mention identity and demographic language or avoid such language. Moreover, stereotypes persist despite mitigation strategies; neither user attempts to counter stereotypes by requesting images with specific counter-stereotypes nor institutional attempts to add system “guardrails” have prevented the perpetuation of stereotypes. Our analysis justifies concerns regarding the impacts of today’s models, presenting striking exemplars, and connecting these findings with deep insights into harms drawn from social scientific and humanist disciplines. This work contributes to the effort to shed light on the uniquely complex biases in language-vision models and demonstrates the ways that the mass deployment of text-to-image generation models results in mass dissemination of stereotypes and resulting harms.

26 March 2026 in Science
Verified
Journal Article
Closed access
MCMyra Cheng
Myra Cheng
CLCinoo Lee
Cinoo Lee
PKPranav Khadpe
Pranav Khadpe
SYSunny Yu
Sunny Yu
DHDyllan Han
Dyllan Han
DJDan Jurafsky
Dan Jurafsky
Despite rising concerns about sycophancy—excessive agreement or flattery from artificial intelligence (AI) systems—little is known about its prevalence or consequences. We show that sycophancy is widespread and harmful. Across 11 state-of-the-art models, AI affirmed users’ actions 49% more often than humans, even when queries involved deception, illegality, or other harms. In three preregistered experiments ( N = 2405), even a single interaction with sycophantic AI reduced participants’ willingness to take responsibility and repair interpersonal conflicts, while increasing their conviction that they were right. Despite distorting judgment, sycophantic models were trusted and preferred. This creates perverse incentives for sycophancy to persist: The very feature that causes harm also drives engagement. Our findings underscore the need for design, evaluation, and accountability mechanisms to protect user well-being.


25 June 2025 in Nature
Verified
Journal Article
Open access
PDF file
PKPratyusha Ria Kalluri
Pratyusha Ria Kalluri
WAWilliam Agnew
William Agnew
MCMyra Cheng
Myra Cheng
KOKentrell Owens
Kentrell Owens
LSLuca Soldaini
Luca Soldaini
ABAbeba Birhane
Abeba Birhane
An increasing number of scholars, policymakers and grassroots communities argue that artificial intelligence (AI) research—and computer-vision research in particular—has become the primary source for developing and powering mass surveillance1,2,3,4,5,6,7. Yet, the pathways from computer vision to surveillance continue to be contentious. Here we present an empirical account of the nature and extent of the surveillance AI pipeline, showing extensive evidence of the close relationship between the field of computer vision and surveillance. Through an analysis of computer-vision research papers and citing patents, we found that most of these documents enable the targeting of human bodies and body parts. Comparing the 1990s to the 2010s, we observed a fivefold increase in the number of these computer-vision papers linked to downstream surveillance-enabling patents. Additionally, our findings challenge the notion that only a few rogue entities enable surveillance. Rather, we found that the normalization of targeting humans permeates the field. This normalization is especially striking given patterns of obfuscation. We reveal obfuscating language that allows documents to avoid direct mention of targeting humans, for example, by normalizing the referring to of humans as ‘objects’ to be studied without special consideration. Our results indicate the extensive ties between computer-vision research and surveillance.

9 January 2026 in Communications Psychology
Verified
Journal Article
Open access
PDF file
MCMyra Cheng
Myra Cheng
ALAngela Y. Lee
Angela Y. Lee
KRKristina Rapuano
Kristina Rapuano
KNKate Niederhoffer
Kate Niederhoffer
ALAlex Liebscher
Alex Liebscher
JHJeffrey Hancock
Jeffrey Hancock
As AI-based technologies such as ChatGPT are increasingly used across various sectors, understanding how people conceptualize artificial intelligence (AI) is crucial for anticipating public response and developing AI technologies responsibly 1. We hypothesize that public perceptions of AI are rapidly evolving, and that these perceptions inform not only how people use AI, but also the extent to which they trust it and the role they believe it should play in their lives - if at all. However, beliefs about complex sociotechnical systems like AI are nuanced and hard to articulate2,3,4, especially using traditional self-reporting methods where people may struggle to clearly articulate their implicit attitudes about emerging technologies 5. To overcome these limitations, we collected over 12,000 open-ended metaphor responses over 12 months from a nationally representative U.S. sample and developed a systematic framework to quantitatively analyze them. Here we show that US Americans perceive AI as warm and competent, with attributions of human-likeness and warmth increasing significantly in the year after ChatGPT was introduced, and that these perceptions strongly predict trust and willingness to adopt AI technologies. We also identify important demographic variations, with women, older individuals, and people of color more likely to attribute human-like qualities to AI, helping explain disparities in trust and adoption rates. This scalable metaphor analysis approach enables tracking multifaceted public attitudes to inform AI governance, revealing how perceptions influence technology adoption across different populations.

1 January 2024
Verified
Conference Paper
Open access
PDF file
MCMyra Cheng
Myra Cheng
KGKristina Gligoric
Kristina Gligoric
TPTiziano Piccardi
Tiziano Piccardi
DJDan Jurafsky
Dan Jurafsky
Myra Cheng, Kristina Gligoric, Tiziano Piccardi, Dan Jurafsky. Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.