Carolyn Penstein Rosé
Professor
- Pittsburgh PA UNITED STATES
Carolyn Rosé's research advances Sociotechnical AI through human-AI collaboration and data-driven inductive biases and representations.
Biography
Her university leadership includes serving as Interim Director of the Language Technologies Institute, Director of the Master of Computational Data Science program, Director of Language Technologies Undergraduate Programs, Faculty Senate member, and Vice Chair of the University Education Council. She launched two highly ranked online certificates in Machine Learning/Data Science and Generative AI, and directed CMU's Generative AI Innovation Incubator in 2023. She has mentored 23 PhD students, over 100 masters students, 9 postdoctoral fellows, and 20+ undergraduates.
Externally, she has held board, editorial, and conference leadership roles across major research societies. In the ACL community, she served as Program Co-Chair for EMNLP 2025, keynote speaker at ACL 2018, and holds editorial and advisory roles in journals and committees. She is a Past President and Inaugural Fellow of the International Society of the Learning Sciences, Founding Chair of the International Alliance to Advance Learning in the Digital Era, Senior IEEE Member, and Executive Editor of the International Journal of Computer-Supported Collaborative Learning (impact factor 6.8). She is also a AAAS Leshner Leadership Institute Fellow (AI Cohort, 2020–2021).
Industry partnerships have spanned Google, Microsoft, Amazon, Oracle, Abridge AI, the Gates Foundation, and the Schmidt Foundation, among others. At the government level, she collaborated for six years with NLP teams at the NIH and Social Security Administration on decision support technology.
Areas of Expertise
Media Appearances
Finalist teams advance in the Amazon Nova AI Challenge: Trusted AI Track
Amazon Science online
2025-06-24
Since November 2024, ten top university teams from around the world have competed in the inaugural Amazon Nova AI Challenge: Trusted AI Track, focused on strengthening security in AI coding assistants and developing new automated methods to ‘red-team’ and test them. After months of intense competition, eight teams have advanced to the finals, demonstrating outstanding innovation in securing AI-powered code generation.
The AI company Elon Musk cofounded just released a 'groundbreaking' tool that can automatically mimic human writing — here's how stunned developers are experimenting with it so far
Business Insider online
2020-07-22
"Historically, natural language generation systems have lacked some nuance," said Carolyn Rose, a professor at Carnegie Mellon University's Language Technologies Institute. But GPT-3 seems different. Based on early reactions, GPT-3 has blown past existing models thanks to its massive dataset and its use of 175 billion parameters — rules the algorithm relies on to decide which word should come next to mimic conversational English. By comparison, the previous version, GPT-2, utilized 1.5 billion parameters, and the next most powerful model — from Microsoft — has 17 billion parameters.
Elon Musk-Backed AI Company Launches New Tool that Writes Naturally Like Humans
Tech Times
2020-07-22
Carnegie Mellon University's Language Technologies Institute Professor Carolyn Rose told Business Insider that while natural language generation systems have historically "lacked some nuance," GPT-3 seems different.
AAAS Selects 28 Mass Media Fellows, Bringing Scientists into Newsrooms Around the Country
American Association for the Advancement of Science online
2020-04-20
The American Association for the Advancement of Science has selected its 2020 Mass Media Science & Engineering Fellows, 28 young scientists who will head to newsrooms around the country this summer for ten weeks of hands-on science reporting. The program places undergraduate, graduate, and post-graduate level scientists, engineers, and mathematicians at media organizations where they write stories for radio and television, newspapers, and magazines.
How Artificial Intelligence Is Changing Teaching
The Chronicle of Higher Education online
2018-08-12
Artificial intelligence is showing up more frequently in college classrooms, particularly at big institutions that are seeking to make large courses more intimate and interactive. A professor at Georgia Tech developed virtual teaching assistants and tutors. Researchers at Carnegie Mellon University are creating conversational agents to promote online discussion. And on a growing number of campuses, professors are using adaptive courseware that adjusts lessons according to students’ understanding and deploying AI-driven tools, like the one Coates used, to promote writing and peer review.
Online Classes Get a Missing Piece: Teamwork
EdSurge online
2016-09-28
Carolyn Rosé, an associate professor in the Human-Computer Interaction Institute at Carnegie Mellon University, has been exploring ways to add social engagement to MOOCs since 2013. She and fellow researchers developed Bazaar, the tool that California community colleges will test in online statistics courses this fall.
How Conversation Impacts Learning
Class Central online
2015-09-16
Prof. Carolyn Rosé’s research focuses on modeling conversations between students in learning contexts to find out what it is about conversations that makes them valuable for learning. To study conversations, she uses text mining, machine learning, and computational discourse analysis. With this new understanding, her goal is to design interventions to support learning in online settings.
Media
Social
Industry Expertise
Education
Carnegie Mellon University:
Ph.D.
Language and Information Technologies
1997
Carnegie Mellon University
M.S.
Computational Linguistics
1994
University of California at Irvine
B.S.
Information and Computer Science
1992
Articles
Adapting to the Long Tail: A Meta-Analysis of Transfer Learning Research for Language Understanding Tasks
Transactions of the Association for Computational Linguistics2022
Natural language understanding (NLU) has made massive progress driven by large benchmarks, but benchmarks often leave a long tail of infrequent phenomena underrepresented. We reflect on the question: Have transfer learning methods sufficiently addressed the poor performance of benchmark-trained models on the long tail? We conceptualize the long tail using macro-level dimensions (underrepresented genres, topics, etc.), and perform a qualitative meta-analysis of 100 representative papers on transfer learning research for NLU. Our analysis asks three questions: (i) Which long tail dimensions do transfer learning studies target? (ii) Which properties of adaptation methods help improve performance on the long tail? (iii) Which methodological gaps have greatest negative impact on long tail performance?
Examining socially shared regulation and shared physiological arousal events with multimodal learning analytics
British Journal of Educational Technology2023
Socially shared regulation contributes to the success of collaborative learning. However, the assessment of socially shared regulation of learning (SSRL) faces several challenges in the effort to increase the understanding of collaborative learning and support outcomes due to the unobservability of the related cognitive and emotional processes. The recent development of trace‐based assessment has enabled innovative opportunities to overcome the problem. Despite the potential of a trace‐based approach to study SSRL, there remains a paucity of evidence on how trace‐based evidence could be captured and utilised to assess and promote SSRL. This study aims to investigate the assessment of electrodermal activities (EDA) data to understand and support SSRL in collaborative learning, hence enhancing learning outcomes
High school students’ data modeling practices and processes: From modeling unstructured data to evaluating automated decisions
Learning, Media and Technology2023
It’s critical to foster artificial intelligence (AI) literacy for high school students, the first generation to grow up surrounded by AI, to understand working mechanism of data-driven AI technologies and critically evaluate automated decisions from predictive models. While efforts have been made to engage youth in understanding AI through developing machine learning models, few provided in-depth insights into the nuanced learning processes. In this study, we examined high school students’ data modeling practices and processes. Twenty-eight students developed machine learning models with text data for classifying negative and positive reviews of ice cream stores. We identified nine data modeling practices that describe students’ processes of model exploration, development, and testing and two themes about evaluating automated decisions from data technologies.
Nine elements for robust collaborative learning analytics: A constructive collaborative critique
International Journal of Computer-Supported Collaborative Learning2023
This editorial represents a collaborative effort between the current co-editors-in-chief of the International Journal of Computer-Supported Collaborative Learning (ijCSCL) and the recent co-editor-in-chief of the Journal of Learning Analytics (JLA), Alyssa Wise, who is also a member of the ijC (LA) have made a presence in ijCSCL. This issue in particular comprises four full articles within this scope, in addition to a timely exposition on Collaborative Learning from an ethics perspective. Thus, it is high time to bring in a voice of leadership from the LA community together with those of the CSCL community to think together about the intersection of work across the two fields.
Examining computational thinking processes in modeling unstructured data
Education and Information Technologies2023
As artificial intelligence (AI) technologies are increasingly pervasive in our daily lives, the need for students to understand the working mechanisms of AI technologies has become more urgent. Data modeling is an activity that has been proposed to engage students in reasoning about the working mechanism of AI technologies. While Computational thinking (CT) has been conceptualized as critical processes that students engage in during data modeling, much remains unexplored regarding how students created features from unstructured data to develop machine learning models. In this study, we examined high school students’ patterns of iterative model development and themes of CT processes in iterative model development. Twenty-eight students from a journalism class engaged in refining machine learning models iteratively for classifying negative and positive reviews of ice cream stores.