Rosalind Picard: Challenges in AI, Wearables, & Smartphones for Understanding and Healing Depression

📅 Jun 05, 2026 Stanford AI for Mental Health 56 MIN 18 SEGMENTS · 2 SPEAKERS

Dr. Rosalind Picard, Sc.D. is the Grover M. Hermann Professor of Health Sciences and Technology at the MIT Media Lab and a ...

What Rosalind Picard said

Written from the verified transcript and checked against it. Every figure links to the moment it was said.

Rosalind Picard discussed her research on using wearables and smartphones to understand and treat depression. She highlighted findings from studies with Mass General Hospital and the Healthy Minds Institute, noting that peripheral signals like electrodermal activity and temperature show changes with depression. In speech analysis, speaking rate and articulation rate were significant predictors of PHQ scores, both within and between individuals. She emphasized the importance of personalized mixed-effects models over traditional machine learning, which often barely beat personal baselines. Picard also described a new app using facial and vocal biomarkers, where positive sentiment and speaking rate were top predictors. She acknowledged limitations, including small datasets and the need to move beyond treating depression as a single PHQ score. Her vision includes predicting individual responses to stressors and preventing relapse using longitudinal data.

Key takeaways

Speaking rate and articulation rate significantly predict PHQ9 scores, both within and between individuals, across two independent studies.
Positive sentiment, speaking rate, and linguistic features were the top SHAP contributors in multimodal depression prediction.
A new app with facial and vocal biomarkers, tested in a nearly 1,000-person study, found significant associations with PHQ8 scores.
Picard aims to predict individual responses to stressors and prevent depression relapse using longitudinal data.

Numbers and commitments

Figure	What it refers to	Type	At
98 seconds	Median daily time spent on speech tasks in the study	metric	33:55
70 people	Sample size of the longitudinal speech study	metric	33:55
12 weeks	Duration of the longitudinal speech study	timeline	33:55
45%	Compliance rate on the speech task	metric	36:48
76%	Proportion of female participants in the Healthy Minds study	metric	38:24
90 seconds	Duration of the video log task	metric	39:33
6 time points	Measurement points in the Healthy Minds study	metric	38:24
8.92	Mean PHQ8 score for a sample participant	metric	46:10

Chapters

0:00Peripheral signals and depression
30:55Machine learning methods and baselines
33:55Speech features and PHQ9 prediction
36:48Study limitations and compliance
38:24Facial and vocal biomarkers app
43:37Acoustic and sentiment findings
46:10Multimodal prediction and SHAP values
50:00Future directions and ethical considerations
53:30Vision for personalized resilience

Asana Delhi 0:00 ↗

Awesome. Thank you for being here. Welcome to the AI for mental health seminar, a monthly series that takes place on the fourth Wednesday of every month. My name is Asana Delhi and I'm hosting this session. I'm an assistant professor in psychiatry and behavioral sciences and by courtesy in biomedical data science and computer science. And for the new faces that I'm seeing here, I want to reiterate that the purpose of AI for MH seminars is first to feature interdisciplinary lectures from experts from this interdisciplinary area. We seek to share cutting-edge AI methods and their clinical or translational applications in psychiatric and neurological care and to raise awareness of AI capabilities among clinicians, researchers, stakeholders and so on in mental health. So we aim for these sessions to foster a vibrant community of clinicians and scientists working together to develop, evaluate and deploy AI systems for mental health care. With that intro, it's really my pleasure and our group's pleasure to host and introduce Dr. Rosalind Picard who is the Governor M. Herman Professor of Health Sciences and Technology at MIT, founder and director of the Affective Computing Research Group at MIT Media Lab, and co-founder of Affectiva and Empatica. Dr. Picard's 1997 book, Affective Computing, helped establish a new branch of computer science, demonstrating why emotion is essential to intelligence and how machines can sense and respond to human affect. Her lab's research has driven advances from emotion-aware interfaces and social assistive technologies to physiological signal sensing in the wild. Through Empatica, her team translated this work into FDA-cleared wearables used to monitor autonomic arousal and support seizure detection, illustrating how rigorous science can become life-changing tools. Her impact has been recognized by election to the National Academy of Engineering and she is a fellow of the IEEE and the ACM. Today she'll discuss progress and future challenges in AI wearables and smartphones for understanding and healing depression. Please join me in giving a warm welcome to Dr. Picard.

1 more exchange in this transcript

Cite this transcript

APA, MLA, BibTeX

APA

Picard, R. (2026, June 5). Rosalind Picard: Challenges in AI, Wearables, & Smartphones for Understanding and Healing Depression [Interview transcript]. Stanford AI for Mental Health. CEOInterviews.AI. https://ceointerviews.ai/interview/964695/

MLA

Rosalind Picard. "Rosalind Picard: Challenges in AI, Wearables, & Smartphones for Understanding and Healing Depression." Stanford AI for Mental Health, 5 Jun. 2026. Transcript, CEOInterviews.AI, https://ceointerviews.ai/interview/964695/.

BibTeX

@misc{picard2026_964695,
  author       = {Rosalind Picard},
  title        = {Rosalind Picard: Challenges in AI, Wearables, \& Smartphones for Understanding and Healing Depression},
  howpublished = {Interview transcript, Stanford AI for Mental Health. CEOInterviews.AI},
  year         = {2026},
  month        = {jun},
  url          = {https://ceointerviews.ai/interview/964695/},
  note         = {Speaker-attributed transcript with timestamps}
}

▶

Key Quotes (4)

“100% of the most dangerous kind of seizure, the grand mall or generalized honoclonic seizures had a more than two standard deviations increase in the pre-Cal in the ectal period over the pre-actal period related to those seizures. So these peaks were real and since then we've got lots more data.”

▶ 10:06

“100 years of literature saying measure the non-dominant side. There are cases where that is going to give the wrong conclusions. When we published this in emotion review, the editor said, you know, it's really interesting data. I suspect this is going to cause a little controversy because they've been doing it this way for about a hundred years.”

▶ 26:54 CONTROVERSIAL

“I think we want to look at like what's going on. Are they experiencing bad sleep? Are they experiencing psycho motor retardation? Are they experiencing these other things? And then I think we can start to tease apart a little bit more of what's going on with some of our data. So we've kind of been limited. We have all this new data we can get, but we're kind of stuck in this old mindset of the gro”

▶ 49:21

“In the future, what I'd really like to do is to be able to tell the difference back here if you're on the red line or the blue line and help you move from the red line to the blue line. So that not if but when those major stressors hit you are able to bounce back.”

▶ 55:07

Mentioned in this interview

Mass General Hospital: Partner in the first wearable ambulatory measurement study on depression.

Healthy Minds Institute: University of Wisconsin team that requested the app for meditation study.

Richie Davidson: Led the Healthy Minds team that collaborated on the meditation app study.

Interspeech: Conference where speech and facial results were presented.

New England Journal of Medicine: Co-edited a series on wearables in clinical practice.

Jeff Ginsburg: Co-editor of the NEJM series on wearables.

Stephen Frenkel: Co-editor of the NEJM series on wearables.

Jeff Drazen: Editor-in-chief who wanted the NEJM series to focus on current practice.

More interviews with Rosalind Picard

Rosalind Picard - Lombardia é Ricerca International Award Winner | National Champion RadioAug 12, 2026 · National Champion Radio · 67 min
Rosalind Picard: Affective Computing, Emotion, Privacy, and HealthAug 03, 2026 · Family Cartoon · 60 min
Civic AI: A tech-builder’s perspective | Professor Rosalind PicardJul 14, 2026 · Institute for Ethics in AI Oxford · 20 min
Is Uncertainty a Creative Superpower? | Rosalind PicardJul 09, 2026 · John Templeton Foundation · 17 min
Rosalind Picard: Affective Computing, Emotion, Privacy, and HealthJun 26, 2026 · MindVoice Production · 60 min
Where AI ends and faith begins, with Rosalind PicardApr 13, 2026 · Thrive Center for Human Development · 51 min

All Rosalind Picard transcripts →

About Rosalind Picard

Rosalind Picard, a professor at MIT and cofounder of Empatica, received the 2026 IEEE Medal for Innovations in Healthcare Technology at the IEEE Honors Ceremony on April 24, 2026, in New York City. The award recognized her pioneering contributions to wearable affective computing for health and well-being. In her acceptance remarks, Picard credited the MIT Media Lab and the Empatica team for developing and bringing to market the first FDA-cleared smartwatch for seizure monitoring, and she described receiving calls from families who said the technology helped save a loved one. In recent public appearances, Picard has discussed the applications and ethical concerns of affective computing. At the Civic AI Conference in July 2026, she presented a technology-builder's perspective on designing AI systems that support human wellbeing, including examples of AI used in healthcare settings. In a July 2026 interview, she criticized what she described as the misuse of chatbot empathy technology by large AI companies, saying it is used to addict people and attach children to technology. She has also called for regulations around data ownership and emotion recognition, stating that people should own their data rather than companies like Amazon or Google, and that existing restrictions on lie detection should be extended to emotion recognition. At a Templeton event in July 2026, she explored the role of uncertainty in creativity and engineering, and in earlier interviews she has discussed the importance of faith in driving scientific inquiry and the need to focus research on problems faced by underserved populations.

All Rosalind Picard appearances → Stanford AI for Mental Health on YouTube →