Karthik Dinakar, Research

My work spans probabilistic graphical models, causal inference, psychiatry, cardiology, neuro-gastroenterology and more recently, computational jurisrudence of the carceral state.

My research interests therefore revolve around these topics:

statistical machine learning (ensemble learning, probabilistic graphical models in particular),
natural language processing (graphical topical models, commonsense reasoning, sociolingusitics & discourse analysis),
human-computer interaction (fail-soft computing, human-in-the-loop computation & data visualization),
causal inference of observational data (Box's loop, posterior predictive checks, potential outcomes approaches),
novel experiment designs (factorial single-subject designs, behavior change and manipulation in the wild),

Selected Projects

Fathom: Computational Empathy

A society of advanced probabilistic graphical models working in tandem with a fail-soft paradigm to create a national crisis text helpline and to bring large scale data science to understanding psychological illnesses. MIT

Ananda

Mixed-iniative variations of learning algorithms that focuses on the embedding of human judgment and insight in the inner inference loops, with a special emphasis on topic models and kernel methods. MIT

Ruminati: ModelingDetection of Textual Cyberbullying

A computationa framework using kernel methods, commonsense reasoning and topic models to detect instances of textual cyberbullying on social media in conjunction with reflective user interaction to combat it. MIT

iBike: Supervised learning for optimizing bike sharing

Supervised machine learning techniques to optimize bike sharing programs by predicting per-hour capacity demand patterns for individual stations in conjunction with real-time visualizations. CMU

Agile Development

Agile development deemphasizes long-term planning in favor of short-term adaptiveness, a strength in a rapidly changing dev environment. However, this short-term focus creates a temptation to neglect best practices for long-term success. How might one find the right balance? CMU

Publications

Calvo, R. A., Dinakar, K., Picard, R., Christensen, H., & Torous, J. (2018). Toward Impactful Collaborations on Computing and Mental Health. Journal of medical Internet research, 20(2), e49. PDF
Barabas, C., Dinakar, K., Virza, M., Ito, J. & Zittrain, J.. (2018). Interventions over Predictions: Reframing the Ethical Debate for Actuarial Risk Assessment. Proceedings of the 1st Conference on Fairness, Accountability and Transparency, in PMLR 81:62-76 PDF
Dinakar, K,., Lensing Machines : representing perspective in machine learning, PhD Thesis, Massachusetts Institute of Technology, 2017. PDF
RA Calvo, K. Dinakar, R. Picard, J. Torous (to appear). "2nd Symposia on Computing and Mental Health." in Proceedings of the 2017 CHI Conference Extended Abstracts on Human Factors in Computing Systems, Denver CO. PDF
R.A. Calvo, K. Dinakar, R. Picard, P. Maes “Computing in Mental Health” To appear in the ACM CHI ’16 Extended Abstracts on Human Factors in Computing Systems. San Jose, CA, May 2016. PDF
Weinstein, E., Thomas, S., Kim, J., White A., Dinakar, K., Selman, R. "How to cope with digital stress: The recommendations adolescents offer their peers online". Journal of Adolescent Research, April 2015. PDF
Dinakar, K., Picard., R, Lieberman, H., "Commonsense Reasoning for Detection, Prevention, and Mitigation of Cyberbullying (Extended Abstract)". , International Joint Conference on Artificial Intelligence, Buenos Aires, Argentina, July 2015. Invited Paper PDF
Dinakar, K., Chen, Jackie., Lieberman, H., Picard., R, "Mixed-Initiative Real-Time Topic Modeling & Visualization for Crisis Counseling". To appear in the Twentieth ACM International Conference on Intelligent User Interfaces, 2015, Atlanta GA. PDF
Dinakar, K., Chaney, J.B.C., Lieberman, H., Blei, D., "Real-time Topic Models for Crisis Counseling". In the Twentieth ACM SIGKDD Conference on Knowledge Discovery and Mining, Workshop on Data Science for the Social Good, 2014, New York, NY. PDF
Dinakar, K., Weinstein, E., Lieberman, H., Selman, R., "Stacked Generalization Learning to Analyze Adolescent Distress". In the AAAI Eigth International Conference on Weblogs and Social Media, 2014, Ann Arbor, Michigan. PDF
Lieberman, H., Dinakar, K., Jones, B., 2013. "Crowdsourced ethics with personalized story matching". In CHI '13 Extended Abstracts on Human Factors in Computing Systems (CHI EA '13). ACM, New York, NY, USA, 709-714. PDF
Dinakar, K., "Ruminati: Modeling the Detection of Textual Cyberbullying", Masters Thesis, Massachusetts Institute of Technology.PDF
Dinakar, K., Jones, B., Lieberman, H., R.,Picard, R.W., Rose, C.P., Thoman, M., Reichart,R., "You Too?! Mixed Initiative LDA story-matching to help teens in distress". In the AAAI Sixth International Conference on Weblogs and Social Media, 2012, Dublin, Ireland. PDF
Dinakar, K., Jones, B., Havasi, C., Lieberman, H., Picard, R., "Commonsense Reasoning for Detection, Prevention, and Mitigation of Cyberbullying.", In the ACM Transactions on Interactive Intelligent Systems, 2011. PDF
Lieberman, H., Dinakar, K., Jones, B., "Let's Gangup on Cyberbullying", IEEE Computer, Social Computing, September 2011.PDF
Dinakar K., Reichart R.,Lieberman, H., "Modeling the detection of textual cyberbullying". AAAI International Conference on Weblog and Social Media - Social Mobile Web Workshop, Barcelona, Spain 2011. PDF
Dinakar K., "Agile development: Overcoming a short-term focus in implementing best practices". ACM SIGPLAN sponsored Proceedings of the Conference on Object-Oriented Programming, Systems, Languages, and Applications (OOPSLA '09), 2009. PDF

Invited Talks & Panels

Latent Variable Models for Understanding Self-Harm, April 16th, 2015 Keynote address tat the 48th conference of the American Association of Suicidology in Atlanta, GA.
Latent Variable Models, January 12th, 2015 William James Hall, Harvard University
Data Visualization, January 13th, 2015 William James Hall, Harvard University
White House summit on sexual assault on college campuses - I demoed a natural language interface powered by probabilistic graphical topic models to scale mental health crisis help architectures in the United States.
Symposium on computational detection of cyberbullying, organized jointly by Delft University, University of Amsterdam & University of Leiden, September 20th, 2013, Leiden, Netherlands.
Common Sense Reasoning for Detection, Prevention and Mitigation of Cyberbullying, International Conference on Intelligent User Interfaces, Santa Monica, March 19- 22, 2013
The White House Conference on Bullying Prevention, March 10th, 2011, Washington D.C – invited to present my research on computational detection of cyberbullying to the special working group on tackling cyberbullying
Computational Empathy – April 3rd, 2012. Graduate School of Education & Department of Psychology, Harvard University.
Cyberbullying: Dodging the Bullet, IEEE Social Computing Conference 2011, 9th October 2011 Cambridge, MA USA
MIT Medical Diversity Dinner – From Blackeye to Blackberry , March 23rd, 2011 – invited to a four-person panel to deliberate ways and means in which MIT can help contribute to tackling bullying
Agile development: Overcoming a short-term focus in implementing best practices. ACM SIGPLAN sponsored Proceedings of the Conference on Object-Oriented Programming, Systems, Languages, and Applications (OOPSLA '09), 2009.

Selected Popular Press Coverage

Secret Codes: Literature By the Numbers

Nautlius Magazine Link
BBC Click: Cyberbullying

BBC World Service Link
Whats Happening in the Mathematical Sciences, Vol9: Thinking Topically

American Mathematical Society Link
AI systems could fight cyberbullying

New Scientist Link
Software Detects and Deters Cyberbullying

Discovery News Link
Delivering on the Promise of Innovation to Help Prevent CyberBullying

The White House Link
Researchers Create Artificial Intelligence To Flag Cyberbullying

Slate Magazine Link