Ongoing research


The Moral Machine web site.

Experimental Ethics of Artificial intelligence

Adoption of self-driving, Autonomous Vehicles (AVs) promises to dramatically reduce the number of traffic accidents. But some inevitable accidents will require AVs to make tradeoffs about potential risk, such as risk to pedestrians on the road versus risk to the passenger in the car. Even if these "moral dilemmas" are rare, defining algorithms to make such decisions is a challenge, since people may be uncomfortable with the idea of Artificial Intelligence making life-and-death decisions without human oversight. Experimental ethics can help manufacturers and regulators understand the psychological challenges that may undermine trust in driverless cars, and our ability to exercise oversight over their behavior. This may be a necessary pre-condition to the wide adoption of autonomous transportation.

Interactive: Moral Machine web site

Seminars: TED Talk

Scientific writings:

Selected Media: New York Times (1)New York Times (2)Washington Post (1)Wall Street JournalTimeIndependent (1)GuardianCBS NewsLA TimesForbesNewsweekCBC (Canada)ABC (Australia),Le Monde (FR)El Pais (ES)ScienceScientific AmericanNew ScientistPBS NOVA NextPBSWiredHuffington PostMIT News,Independent (2)Washington Post (2)Washington Post (3)BBC World Service (live interview)Huffington Post (live interview)New York MagazinePopular ScienceMIT Technology Review


human-machine cooperation

Since Alan Turing envisioned Artificial Intelligence (AI), a major driving force behind technical progress has been competition with human cognition (e.g. beating humans in Chess or Jeopardy!). Less attention has been given to developing autonomous machines that learn to cooperate with humans. Cooperation does not require sheer computational power, but relies on intuition, and pre-evolved dispositions toward cooperation, common-sense mechanisms that are difficult to encode in machines. We develop state-of-the-art machine-learning algorithms that cooperate with people and other machines at levels that rival human cooperation in two-player repeated games. 

Scientific writings:

  • J. W. Crandall, M. Oudah, Tennom, F. Ishowo-Oloko, S. Abdallah, J.-F. Bonnefon, M. Cebrian, A. Shariff, M. A. Goodrich,, I. Rahwan. Cooperating with MachinesarXiv:arXiv:1703.06207v2 [cs.AI]

Selected Media:  ScienceThe RegisterLa RepubblicaMIT Technology Review

future of work: automation and labor

As advances in robotics and artificial intelligence revive concerns about the impact of automation on jobs, a question looms: How will automation affect employment in different cities and economies? We use tools from complex systems and urban science to explore this question.

Scientific Writings:

The Nightmare Machine web site.

The Nightmare Machine

Since centuries, and across geographies, religions, and cultures, people have tried to innovate ways of scaring each other. Creating a visceral emotion such as fear remains one of the cornerstones of human creativity. This challenge is especially important in an age in which we wonder what the limits of artificial intelligence are -- in this case, can machines learn to scare us? Towards this goal, we present Haunted Faces and Haunted Places: computer-generated scary imagery powered by deep learning algorithms. 

Interactive: Nightmare Machine web site

Selected Media: Washington PostThe AtlanticForbesBBCNPR,CNETNBC NewsVice


Can we improve the ability of machines to understand human emotions? We use Deep Learning techniques, with a novel pre-training approach using emojis, to outperform many state-of-the-art techniques in sentiment, emotion and sarcasm detection. 

Interactive: DeepMoji web site

Code: DeepMoji pre-trained model on GitHub

Scientific writings:

Selected Media: MIT Technology Review, The Telegraph, BBC, Newsweek, The Register, Digital Trends


Dynamics of Political Development

Political constitutions describe the fundamental principles by which nation-states are governed, the political and legal state institutions, the powers, procedures, and duties of those institutions, and the rights and responsibilities of individuals. How do these constitutions develop over long periods of time? and what is the interplay between colonial history and global, time varying trends in determining the characteristics of a country's constitution? We explore these questions using new techniques of computational social science.

Scientific writings:

society-in-the-loop: Algorithmic Social Contract

Recent rapid advances in Artificial Intelligence (AI) and Machine Learning have raised many questions about the regulatory and governance mechanisms for autonomous machines. This is not about individual gadgets, but about complex, networked systems of humans and algorithms making decisions in business, government and the media. We need conceptual frameworks for designing new governance architectures for these human-machine social systems. In doing so, it is helpful to learn lessons about human cooperation and governance from political philosophy and cultural anthropology. It is also important to understand the limits of regulation, and whether over-regulation of adaptive systems may even backfire.

Scientific writings:

Selected Media: New York Times.


Past research


crowdsourcing a manhunt

People often say that we live in a small world. In a brilliant experiment, legendary social psychologist Stanley Milgram proved the six degrees of separationhypothesis: that everyone is six or fewer steps away, by way of introduction, from any other person in the world. But how far are we, in time, from anyone on earth? Our team won the Tag Challenge, showing it is possible to find a person, using only his or her mug shot, within 12 hours.

Scientific writings:

Selected Media: New ScientistNatureScientific American,EconomistNew ScientistNextgovMIT Technology Review

Time-critical social mobilization: darpa red balloon challenge

In 2009, DARPA launched the Network Challenge, to explore the roles the Internet and social networking play in the timely communication, wide-area team-building, and urgent mobilization required to solve broad-scope, time-critical problems. The challenge was to be the first to locate 10 moored, 8-foot, red weather balloons at 10 random locations in the continental United States. A team from MIT won by locating all balloons in under 9 hours. We helped analyze the factors behind the team's success. We then quantified the limits of this kind of mobilization, and introduced techniques for improving information verification in mass collaboration. 

Scientific writings:

  • A. Rutherford, M. Cebrian, S. Dsouza, E. Moro, A. Pentland, and I. Rahwan (2013). Limits of Social MobilizationProceedings of the National Academy of Sciences, vol. 110 no. 16 pp. 6281-6286
  • G. Pickard, W. Pan, I. Rahwan, M. Cebrian, R. Crane, A. Madan, A. Pentland (2011). Time-Critical Social MobilizationScience. Vol. 334 no. 6055 pp. 509-512.
  • V. Naroditskiy, I. Rahwan, M. Cebrian, N. R. Jennings (2012). Verification in Referral-Based CrowdsourcingPLOS ONE 7(10): e45924.
  • H. Chen, I. Rahwan, and M. Cebrian (2016). Bandit strategies in social search: the case of the DARPA red balloon challengeEPJ Data Science, 2016 5:20

Selected Media: MSNBCPopular MechanicsScience NewsNBC NewsABCLa RepubblicaACMRed OrbitScience 2.0Science DailyMIT News

DARPA Shredder Challenge

The Internet has unleashed the capacity for planetary-scale collective problem solving (also known as crowdsourcing). However, the very openness of crowdsourcing makes it vulnerable to sabotage by rogue or competitive actors. To explore the effect of errors and sabotage on the performance of crowdsourcing, we analyze data from the DARPA Shredder Challenge, a prize competition for exploring methods to reconstruct documents shredded by a variety of paper shredding techniques. 

Scientific writings:

  • N. Stefanovitch, A. Alshamsi, M. Cebrian, I. Rahwan (2014). Error and attack tolerance of collective problem solving: The DARPA Shredder ChallengeEPJ Data Science. vol 3, no 13, pages 1-27.

Selected Media: Nautilus, WiredUCSD Press Release

emotions & Influence in Social Networks

Social networks shape our mood, emotions, and behavior. New unobtrusive sensing techniques (wearable sensors, mobile phones, social media), allow us to study these social dynamics in the real world with unprecedented detail, complementing old methods like lab studies and surveys. We discovered that while some behaviors spread on social networks like a virus (you catch the blues from your friends), other behaviors work in the opposite direction, where our behavior changes to complement the behavior of others. Moreover, our individual personality traits affect how these processes work. These findings inform potential interventions designed to improve societal well-being. 

Scientific writings:

  • A. Alshamsi, F. Pianesi, B. Lepri, A. Pentland, and I. Rahwan (2016). Network Diversity and Affect Dynamics: The Role of Personality TraitsPLOS ONE, 11(4), e0152358.
  • A. Alshamsi, F. Pianesi, B. Lepri, A. Pentland, and I. Rahwan (2015). Beyond Contagion: Reality Mining Reveals Complex Patterns of Social InfluencePLOS ONE 10(8), e0135740.
  • A. Alshamsi, E. Awad, M. Almehrezi, V. Babushkin, P.J. Chang, Z. Shoroye, A.P. Tóth, and I. Rahwan (2015). Misery loves company: happiness and communication in the cityEPJ Data Science. 4(1), pp.1-12.

Cognitive Limits of Social Networks

There is a wide cultural belief in the power of the Web and social media as enablers of collective intelligence. They help us spread information rapidly, and learn useful information from each other. But there are fundamental limits to the capabilities of those networks. Understanding these limits is essential to improving social media and allowing society to make the most of it. 

Scientific writings:

  • M. Cebrian, I. Rahwan, A. Pentland (2016). Beyond ViralCommunications of the ACM. 59(4):36-39. 
  • I. Rahwan, D. Krasnoshtan, A. Shariff, J. F. Bonnefon (2014). Analytical reasoning task reveals limits of social learning in networksJournal of the Royal Society Interface. 11(93).

Selected Media: Royal SocietyDaily MailLive

Promoting Cooperation

How to best govern society and promote cooperation is a centuries-old debate: is cooperation best maintained by a central authority, or is it better handled by more decentralized forms of governance? Using mathematical models, we explore the tradeoffs between different mechanisms for cooperation. We studied how centralized cooperation-enforcement mechanisms may be undermined by corruption, and how decentralized enforcement can provide viable alternatives. We also showed that it is possible to promote cooperation by inducing peer-pressure in social networks, using interventions that amplify the power of peer enforcement mechanisms.

Scientific writings:

  • A. Mani, I. Rahwan, and A. Pentland (2013). Inducing Peer Pressure to Promote CooperationScientific Reports. 3(1735) doi:10.1038/srep01735.
  • S. Abdallah, R. Sayed, I. Rahwan, B. LeVeck, M. Cebrian, A. Rutherford, J. Fowler (2014). Corruption Drives the Emergence of Civil SocietyJournal of the Royal Society Interface. 11(93).

Selected Media: Anti-Corruption Research Network (part of Transparency International)

Mass-Scale Argumentation

This research program aimed to lay the foundations for a world-wide web of annotated arguments, which would allow people to navigate arguments all over the Web. For instance, if you are unsure of the assumptions behind a statement made in an article or piece of news, you would be able to navigate evidence supporting it, as well as the counter-arguments that undermine it, even if these are located elsewhere on the Internet. We articulated this vision, and proposed some semantic annotation technologies that would facilitate it. Building on these ideas, others have grown a full research program to realize this "worldwide argument web" (see this recent survey for an overview).

Scientific writings:

  • I. Rahwan, B. Banihashemi, C. Reed, D. Walton and S. Abdallah (2011). Representing and Classifying Arguments on the Semantic WebThe Knowledge Engineering Review. Volume 26, Issue 4, pp 487-511
  • I. Rahwan (2008). Mass Argumentation and the Semantic WebJournal of Web Semantics. Vol 6, No 1, pages 29-37. 
  • I. Rahwan, F. Zablith and C. Reed (2007). Laying the Foundations for a World Wide Argument WebArtificial Intelligence, Vol 171, No 10-15, pages 897-921
  • I. Rahwan, F. Zablith, and C. Reed (2007). Towards Large Scale Argumentation Support on the Semantic Web. In: Proceedings of 22nd Conference on Artificial Intelligence (AAAI). AAAI Press, California, USA, pages 1446-1451. 
  • I. Rahwan, P. McBurney (2007). Guest Editors' Introduction: Argumentation TechnologyIEEE Intelligent Systems, November/December, Vol. 22, no. 6, pages 21-23. 
  • C. Chesñevar, J. McGinnis, S. Modgil, I. Rahwan, C. Reed, G. Simari, M. South, G. Vreeswijk and S. Willmott (2006). Towards an Argument Interchange FormatThe Knowledge Engineering Review, Vol 21, No 4, pages 293-316. 

Opinion aggregation

Opinion aggregation on social media uses various mechanisms,  such as "Likes" or thumbs-up/-down, which handle a single item at a time. In many domains (e.g., political discussion or AI ethics), we need to consider a very large number of possibilities, or to take into account the relationships between different claims and how they rebut one another through complex webs of arguments and counter-arguments. We study methods for aggregating opinions about such complex domains, the quality of the outcomes of different opinion aggregations methods, and whether strategic agents can manipulate those outcomes.

Scientific writings:

  • R. Noothigattu, S. S. Gaikwad, E. Awad, S. Dsouza, I. Rahwan, P. Ravikumar, A. D. Procaccia (2017). A Voting-Based System for Ethical Decision Making. arXiv:1709.06692 [cs.AI]
  • M. R. Frank, M. Cebrian, G. Pickard, I. Rahwan (2017). Validating Bayesian truth serum in large-scale online human experimentsPLOS ONE. 12(5): e0177385.
  • E. Awad, J.-F. Bonnefon, M. Caminada, T. Malone, I. Rahwan (2017). Experimental Assessment of Aggregation Principles in Argumentation-enabled Collective IntelligenceACM Transactions on Internet Technology. 17:3.
    [Paper] [arXiv preprint]
  • E. Awad, M. Caminada, G. Pigozzi, M. Podlaszweski, I. Rahwan (2017). Pareto Optimality and Strategy Proofness in Group Argument EvaluationJournal of Logic and Computation. (in press)
    [Paper] [Free link to access article]
  • E. Awad, R. Booth, F. Tohme, I. Rahwan (2017). Judgment Aggregation in Multi-Agent ArgumentationJournal of Logic and Computation. 27(1): 227-259.
  • R. Booth, E. Awad, and I. Rahwan (2014). Interval Methods for Judgment Aggregation in Argumentation. In: Proc. 14th Int. Conference on Principles of Knowledge Representation and Reasoning (KR), Vienna.
  • R. Booth, M. Caminada, M. Podlaszewski, and I. Rahwan (2012). Quantifying Disagreement in Argument-based Reasoning. In: Proc. 11th Int. Conference on Autonomous Agents and Multi-Agent Systems (AAMAS), Valencia, Spain.
  • I. Rahwan and K. Larson (2011). Logical Mechanism DesignThe Knowledge Engineering Review. Volume 26, No 1, pages 61-69.
  • S. Pan, K. Larson and I. Rahwan (2010). Argumentation Mechanism Design for Preferred Semantics. In: Proceedings of the 3rd International Conference on Computational Models of Argument (COMMA), Italy, pages 403-414.
  • I. Rahwan and F. Tohmé (2010). Collective Argument Evaluation as Judgement Aggregation. In: Proceedings of the 9th International Joint Conference on Autonomous Agents and Multiagent Systems (AAMAS), Toronto, Canada.
  • I. Rahwan, K. Larson and F. Tohmé (2009). A Characterisation of Strategy-Proofness for Grounded Argumentation Semantics. In: Proceedings of the 21st International Joint Conference on Artificial Intelligence (IJCAI), Pasadena, California, USA.
  • I. Rahwan and K. Larson (2008). Pareto Optimality in Abstract Argumentation. In: Proceedings of 23rd Conference on Artificial Intelligence (AAAI). AAAI Press, California, USA, pages 150-155.
  • I. Rahwan and K. Larson (2008). Mechanism Design for Abstract Argumentation. In: Proceedings of 7th International Conference on Autonomous Agents and Multiagent Systems (AAMAS). Estoril, Portugal, pages 1031-1038. 

Computational models of argumentation & negotiation

Computer scientists are interested in developing models of argumentation and negotiation processes, in order to facilitate conflict-resolution and deliberation among machines and between humans and machines. These models are often formalized using formal logical theories, or computer simulation, and can be tested using human subject experiments.

Scientific writings:

  • S. Dsouza, Y. Gal, P. Pasquier, S. Abdallah, and I. Rahwan (2013). Reasoning about Goal Revelation in Human NegotiationIEEE Intelligent Systems. vol. 28, no. 2, pp. 74-80.
  • P. Pasquier, R. Hollands, I. Rahwan, F. Dignum and L. Sonenberg (2011). An Empirical Study of Interest-based NegotiationAutonomous Agents and Multiagent Systems. Volume 22, Number 2, 249-288
  • I. Rahwan, M. I. Madakkatel, J. F. Bonnefon, R. N. Awan and S. Abdallah (2010). Behavioural Experiments for Assessing the Abstract Argumentation Semantics of ReinstatementCognitive Science. 34(8):1483–1502
  • F. Grasso, I. Rahwan, C. Reed, and G. R. Simari (2010) Introducing Argument & ComputationArgument & Computation, Vol 1, No 1, pages 1-5.
  • A. Belesiotis, M. Rovatsos and I. Rahwan (2010). Agreeing on Plans through Iterated Disputes. In: Proceedings of the 9th International Joint Conference on Autonomous Agents and Multiagent Systems (AAMAS), Toronto, Canada.
  • N. C. Karunatillake, N. R. Jennings, I. Rahwan and P. McBurney (2009). Dialogue Games that Agents Play within a SocietyArtificial Intelligence. Vol 173, No 9-10, pages 935-981.
  • I. Rahwan, P. Pasquier, L. Sonenberg and F. Dignum (2009). A Formal Analysis of Interest-based NegotiationAnnals of Mathematics and Artificial Intelligence. Vol 55, Issue 3-4, pages 253- 276.
  • P. Pasquier, L. Sonenberg, I. Rahwan, F. Dignum, R. Hollands (2007). An Empirical Study of Interest-based Negotiation. In: Proceedings of 9th International Conference on Electronic Commerce (ICEC). ACM Press.
  • I. Rahwan, L. Sonenberg, N. R. Jennings and P. McBurney (2007). STRATUM: A Methodology for Designing Heuristic Agent Negotiation StrategiesApplied Artificial Intelligence, Vol 21, No 6, pages 489-527. 
  • I. Rahwan, L. Amgoud (2006). An Argumentation-based Approach for Practical Reasoning. Proceedings of the 5th International Joint Conference on Autonomous Agents and Multiagent Systems (AAMAS), Hakodate, Japan
  • N. C. Karunatillake, N. R. Jennings, I. Rahwan and S. D. Ramchurn (2006). Managing Social Influences through Argumentation-based Negotiation. Proceedings of the 5th International Joint Conference on Autonomous Agents and Multiagent Systems (AAMAS), Hakodate, Japan
  • P. Pasquier, I. Rahwan, F. Dignum and L. Sonenberg (2006). Argumentation and Persuasion in the Cognitive Coherence Theory. In P. Dunne and T. Bench-Capon (Eds.), Proceedings of the 1st International Conference on Computational Models of Argument (COMMA), IOS Press, Amsterdam. 
  • N. C. Karunatillake, N. R. Jennings, I. Rahwan and T. Norman (2005). Argument-based Negotiation in a Social Context. Proceedings of the International Conference on Autonomous Agents and Multiagent Systems (AAMAS)
  • I. Rahwan, F. Koch, C. Graham, J. A. Kattan and L. Sonenberg (2005). Goal-directed Automated Negotiation for Supporting Mobile User Coordination. In  Modeling and Using Context: 5th International and Interdisciplinary Conference, CONTEXT, LNCS, Vol 3554, Springer
  • I. Rahwan, L. Sonenberg and F. Dignum (2003). Towards Interest-Based Negotiation. Proceedings of the 2nd International Conference on Autonomous Agents and Multi-Agent Systems (AAMAS), Melbourne, Australia
  • I. Rahwan, S. D. Ramchurn, N. R. Jennings, P. McBurney, S. Parsons and L. Sonenberg (2003). Argumentation-Based NegotiationThe Knowledge Engineering Review, Volume 18, No. 4, pages 343-375.