Research Publications

Universals and Variations in Moral Decisions Made in 42 Countries by 70,000 Participants (2020)

https://doi.org/10.1073/pnas.1911517117

Authors: Edmond Awad, Sohan Dsouza, Azim Shariff, Iyad Rahwan, Jean-François Bonnefon
Publication date: 2020/2/4
Journal: Proceedings of the National Academy of Sciences
Volume: 117
Issue: 5
Pages: 2332-2337
Publisher: National Academy of Sciences

We report the largest cross-cultural study of moral preferences in sacrificial dilemmas, that is, the circumstances under which people find it acceptable to sacrifice one life to save several. On the basis of 70,000 responses to three dilemmas, collected in 10 languages and 42 countries, we document a universal qualitative pattern of preferences together with substantial country-level variations in the strength of these preferences. In particular, we document a strong association between low relational mobility (where people are more cautious about not alienating their current social partners) and the tendency to reject sacrifices for the greater good—which may be explained by the positive social signal sent by such a rejection. We make our dataset publicly available for researchers.

Drivers are Blamed More than their Automated Cars when Both Make Mistakes (2020)

https://doi.org/10.1038/s41562-019-0762-8

Authors: Edmond Awad, Sydney Levine, Max Kleiman-Weiner, Sohan Dsouza, Joshua B Tenenbaum, Azim Shariff, Jean-François Bonnefon, Iyad Rahwan
Publication date: 2020/2
Journal: Nature Human Behaviour
Volume: 4
Issue: 2
Pages: 134-143
Publisher: Nature Publishing Group

When an automated car harms someone, who is blamed by those who hear about it? Here we asked human participants to consider hypothetical cases in which a pedestrian was killed by a car operated under shared control of a primary and a secondary driver and to indicate how blame should be allocated. We find that when only one driver makes an error, that driver is blamed more regardless of whether that driver is a machine or a human. However, when both drivers make errors in cases of human–machine shared-control vehicles, the blame attributed to the machine is reduced. This finding portends a public under-reaction to the malfunctioning artificial intelligence components of automated cars and therefore has a direct policy implication: allowing the de facto standards for shared-control vehicles to be established in courts by the jury system could fail to properly regulate the safety of those vehicles; instead, a top-down scheme (through federal laws) may be called for.

Crowdsourcing Moral Machines (2020)

https://doi.org/10.1145/3339904

Authors: Edmond Awad, Sohan Dsouza, Jean-François Bonnefon, Azim Shariff, Iyad Rahwan
Publication date: 2020/2/24
Journal: Communications of the ACM
Volume: 63
Issue: 3
Pages: 48-55
Publisher: ACM

A platform for creating a crowdsourced picture of human opinions on how machines should handle moral dilemmas.

A Computational Model of Commonsense Moral Decision Making (2018)

https://doi.org/10.1145/3278721.3278770

Authors: Richard Kim, Max Kleiman-Weiner, Andrés Abeliuk, Edmond Awad, Sohan Dsouza, Joshua B Tenenbaum, Iyad Rahwan
Publication date: 2018/12/27
Conference: Proceedings of the 2018 AAAI/ACM Conference on AI, Ethics, and Society
Pages: 197-203
Publisher: ACM

We introduce a computational model for building moral autonomous vehicles by learning and generalizing from human moral judgments. We draw on a cognitively inspired model of how people and young children learn moral theories from sparse and noisy data and integrate observations made from different people in different groups. The problem of moral learning for autonomous vehicles is cast as learning how to weigh the different features of the dilemma using utility calculus, with the goal of making these trade-offs reflect how people make them in a wide variety of moral dilemma. By modeling the structures of individuals and groups in a hierarchical Bayesian model, we show that an individual's moral values -- as well as a group's shared values -- can be inferred from sparse and noisy data. We evaluate our approach with data from the Moral Machine, a web application that collects human judgments on moral dilemmas involving autonomous vehicles, and show that the model rapidly and accurately infers people's preferences and can predict the difficulty of moral dilemmas from limited data.

The Moral Machine Experiment (2018)

https://doi.org/10.1038/s41586-018-0637-6

Authors: Edmond Awad, Sohan Dsouza, Richard Kim, Jonathan Schulz, Joseph Henrich, Azim Shariff, Jean-François Bonnefon, Iyad Rahwan
Publication date: 2018/11
Journal: Nature
Volume: 563
Issue: 7729
Pages: 59-64
Publisher: Nature Publishing Group

With the rapid development of artificial intelligence have come concerns about how machines will make moral decisions, and the major challenge of quantifying societal expectations about the ethical principles that should guide machine behaviour. To address this challenge, we deployed the Moral Machine, an online experimental platform designed to explore the moral dilemmas faced by autonomous vehicles. This platform gathered 40 million decisions in ten languages from millions of people in 233 countries and territories. Here we describe the results of this experiment. First, we summarize global moral preferences. Second, we document individual variations in preferences, based on respondents’ demographics. Third, we report cross-cultural ethical variation, and uncover three major clusters of countries. Fourth, we show that these differences correlate with modern institutions and deep cultural traits. We discuss how these preferences can contribute to developing global, socially acceptable principles for machine ethics. All data used in this article are publicly available.

[Selected Media: The New Yorker, Washington Post, The Economist, BBC, Nature News, MIT Technology Review, Fast Company, Motherboard / Vice, Business Insider, The Guardian, Scientific American, WIRED, The Verge, Spiegel, Le Monde, Prospect ]

A Voting-Based System for Ethical Decision Making (2018)

https://doi.org/10.1609/aaai.v32i1.11512

Authors: Ritesh Noothigattu, Snehalkumar Gaikwad, Edmond Awad, Sohan Dsouza, Iyad Rahwan, Pradeep Ravikumar, Ariel Procaccia
Publication date: 2018/4/25
Journal: Proceedings of the AAAI Conference on Artificial Intelligence
Volume: 32
Issue: 1

We present a general approach to automating ethical decisions, drawing on machine learning and computational social choice. In a nutshell, we propose to learn a model of societal preferences, and, when faced with a specific ethical dilemma at runtime, efficiently aggregate those preferences to identify a desirable choice. We provide a concrete algorithm that instantiates our approach; some of its crucial steps are informed by a new theory of swap-dominance efficient voting rules. Finally, we implement and evaluate a system for ethical decision making in the autonomous vehicle domain, using preference data collected from 1.3 million people through the Moral Machine website.

A Participatory Sensing Approach for Personalized Distance-to-Empty Prediction and Green Telematics (2015)

https://doi.org/10.1145/2768510.2768530

Authors: Chien-Ming Tseng, Chi-Kin Chau, Sohan Dsouza, Erik Wilhelm
Publication date: 2015/7/14
Conference: Proceedings of the 2015 ACM Sixth International Conference on Future Energy Systems
Pages: 47-56
Publisher: ACM

Participatory sensing is an emerging concept that integrates crowd-sourced data collection and knowledge discovery of collective behavior. Capitalizing on the advent of abundant sensors and information collection systems in near-future vehicles, we develop a participatory sensing based system and its methodologies for driving energy efficiency applications. Distance-to-empty (DTE) is the distance an electric or internal-combustion engine (ICE) vehicle can reach before its energy/fuel is exhausted, which is determined by a variety of uncertain factors, such as driving behavior, terrain, types of road, traffic, and vehicle specification. Green telematics aims to optimize the route selection with lower energy consumption. In this paper, we explore an effective approach that integrates the vehicle data gathered from participatory sensing to provide more accurate personalized DTE prediction and green telematics. Our approach relies on extracting the driver/vehicle/route dependent features and discovering correlations from collective driving data. We also present concrete case studies of our results, such as (1) DTE prediction for EVs based on the data of ICE vehicles, (2) classification and recommendations of energy-efficient driving behavior, and (3) route-level energy consumption geo-fencing and planning.

Cloudthink: a Scalable Secure Platform for Mirroring Transportation Systems in the Cloud (2015)

https://doi.org/10.3846/16484142.2015.1079237

Authors: Erik Wilhelm, Joshua Siegel, Simon Mayer, Leyna Sadamori, Sohan Dsouza, Chi-Kin Chau, Sanjay Sarma
Publication date: 2015/7/3
Journal: Transport
Volume: 30
Issue: 3
Pages: 320-329
Publisher: Taylor & Francis

We present a novel approach to developing a vehicle communication platform consisting of a low-cost, open-source hardware for moving vehicle data to a secure server, a Web Application Programming Interface (API) for the provision of third-party services, and an intuitive user dashboard for access control and service distribution. The CloudThink infrastructure promotes the commoditization of vehicle telematics data by facilitating easier, flexible, and more secure access. It enables drivers to confidently share their vehicle information across multiple applications to improve the transportation experience for all stakeholders, as well as to potentially monetize their data. The foundations for an application ecosystem have been developed which, taken together with the fair value for driving data and low barriers to entry, will drive adoption of CloudThink as the standard method for projecting physical vehicles into the cloud. The application space initially consists of a few fundamental and important applications (vehicle tethering and remote diagnostics, road-safety monitoring, and fuel economy analysis) but as CloudThink begins to gain widespread adoption, the multiplexing of applications on the same data structure and set will accelerate its adoption.

[ Selected Media: Masdar Institute News ]

A Social Approach for Predicting Distance-to-Empty in Vehicles (2014)

https://doi.org/10.1145/2602044.2602073

Authors: Chien-Ming Tseng, Sohan Dsouza, Chi-Kin Chau
Publication date: 2014/6/11
Conference: Proceedings of the 5th international conference on Future energy systems
Pages: 215-216
Publisher: ACM

Distance-to-Empty (DTE) in vehicles depends on several uncertain factors, such as speed, terrain, traffic and driving behavior. Accurate estimation of DTE is vital for not only the scheduling for refueling, but also for the choice of routes for the budget- and/or environmentally-conscious. Traditional approaches often rely on a single driver's personal history. In this paper, we explore a social approach by using other drivers' data to predict the fuel consumption for a given driver along a new route that is not traveled previously. We develop a least-squares regression model and corroborate the performance empirically by an on-road, multi-driver experiment. Our results can enable a new kind of social platform for trip planning based on the shared data among drivers.

Targeted Social Mobilization in a Global Manhunt (2013)

https://doi.org/10.1371/journal.pone.0074628

Authors: Alex Rutherford, Manuel Cebrian, Iyad Rahwan, Sohan Dsouza, James McInerney, Victor Naroditskiy, Matteo Venanzi, J. R. Jennings, deLara, Nicholas R, Eero Wahlstedt, Steven U Miller
Publication date: 2013/9/30
Journal: PloS one
Volume: 8
Issue: 9
Pages: e74628
Publisher: Public Library of Science

Social mobilization, the ability to mobilize large numbers of people via social networks to achieve highly distributed tasks, has received significant attention in recent times. This growing capability, facilitated by modern communication technology, is highly relevant to endeavors which require the search for individuals that possess rare information or skills, such as finding medical doctors during disasters, or searching for missing people. An open question remains, as to whether in time-critical situations, people are able to recruit in a targeted manner, or whether they resort to so-called blind search, recruiting as many acquaintances as possible via broadcast communication. To explore this question, we examine data from our recent success in the U.S. State Department's Tag Challenge, which required locating and photographing 5 target persons in 5 different cities in the United States and Europe – in under 12 hours – based only on a single mug-shot. We find that people are able to consistently route information in a targeted fashion even under increasing time pressure. We derive an analytical model for social-media fueled global mobilization and use it to quantify the extent to which people were targeting their peers during recruitment. Our model estimates that approximately 1 in 3 messages were of targeted fashion during the most time-sensitive period of the challenge. This is a novel observation at such short temporal scales, and calls for opportunities for devising viral incentive schemes that provide distance or time-sensitive rewards to approach the target geography more rapidly. This observation of ′12 hours of separation' between individuals has applications in multiple areas from emergency preparedness, to political mobilization.

[Selected Media: MIT Technology Review, New Scientist, Nature]

Reasoning about Goal Revelation in Human Negotiation (2013) 

https://doi.org/10.1109/MIS.2011.93

Authors: Sohan Dsouza, Ya’akov Gal, Philippe Pasquier, Sherief Abdallah, Iyad Rahwan
Publication date: 2013/4/4
Journal: Intelligent Systems
Volume: 28
Issue: 2
Pages: 74-80
Publisher: IEEE

This article studies how people reveal private information in strategic settings in which participants need to negotiate over resources but are uncertain about each other's objectives. The study compares two negotiation protocols that differ in whether they allow participants to disclose their objectives in a repeated negotiation setting of incomplete information. Results show that most people agree to reveal their goals when asked, and this leads participants to more beneficial agreements. Machine learning was used to model the likelihood that people reveal their goals in negotiation, and this model was used to make goal request decisions in the game. In simulation, use of this model is shown to outperform people making the same type of decisions. These results demonstrate the benefit of this approach towards designing agents to negotiate with people under incomplete information.

Limits of Social Mobilization (2013)

https://doi.org/10.1073/pnas.1216338110

Authors: Alex Rutherford, Manuel Cebrian, Sohan Dsouza, Esteban Moro, Alex Pentland, Iyad Rahwan
Publication date: 2013/4/16
Journal: Proceedings of the National Academy of Sciences
Volume: 110
Issue: 16
Pages: 6281-6286
Publisher: National Academy of Sciences

The Internet and social media have enabled the mobilization of large crowds to achieve time-critical feats, ranging from mapping crises in real time, to organizing mass rallies, to conducting search-and-rescue operations over large geographies. Despite significant success, selection bias may lead to inflated expectations of the efficacy of social mobilization for these tasks. What are the limits of social mobilization, and how reliable is it in operating at these limits? We build on recent results on the spatiotemporal structure of social and information networks to elucidate the constraints they pose on social mobilization. We use the DARPA Network Challenge as our working scenario, in which social media were used to locate 10 balloons across the United States. We conduct high-resolution simulations for referral-based crowdsourcing and obtain a statistical characterization of the population recruited, geography covered, and time to completion. Our results demonstrate that the outcome is plausible without the presence of mass media but lies at the limit of what time-critical social mobilization can achieve. Success relies critically on highly connected individuals willing to mobilize people in distant locations, overcoming the local trapping of diffusion in highly dense areas. However, even under these highly favorable conditions, the risk of unsuccessful search remains significant. These findings have implications for the design of better incentive schemes for social mobilization. They also call for caution in estimating the reliability of this capability.

[ Selected Media: NBC News, ABC, Arabian Gazette, Gulf Today

Global Manhunt Pushes the Limits of Social Mobilization (2013)

https://doi.org/10.1109/MC.2012.295

Authors: Iyad Rahwan, Sohan Dsouza, Alex Rutherford, Victor Naroditskiy, James McInerney, Matteo Venanzi, Nicholas Jennings, Manuel Cebrian
Publication date: 2013/4/1
Journal: Computer
Volume: 46
Issue: 4
Pages: 68-75
Publisher: IEEE

Using social media and only the targets' mug shots, a team competing in the US State Department-sponsored Tag Challenge located three of five targeted people in five cities in the US and Europe in less than 12 hours.

[ Selected Media: The Economist, Nextgov, New Scientist, Popular Science, The National, Vision (cover story) ]

The Effects of Goal Revelation on Computer-Mediated Negotiation (2009)

https://escholarship.org/uc/item/7g54k0vt

Authors: Ya’akov Gal, Sohan D’souza, Philippe Pasquier, Iyad Rahwan, Sherief Abdallah

Publication date: 2009

Journal: Proceedings of the Annual meeting of the Cognitive Science Society (CogSci), Amsterdam, The Netherlands

This paper studies a novel negotiation protocol in settings in which players need to exchange resources in order to achieve their own objective, but are uncertain about the objectives of other participants. The protocol allows participants to request each other to disclose their interests at given points in the negotiation. Revealing information about participants’ needs may facilitate agreement, but it also exposes their negotiation strategy to the exploitation of others. Empirical studies were conducted using computer-mediated negotiation scenarios that provided an analogue to the way goals and resources interact in the world. The scenarios varied in the individual positions and interests of participants, as well as the dependency relation- ships that hold between participants. Results show that those who choose to reveal their underlying goals outperform nego- tiators in the same setting that use a protocol that forbids reve- lation. In addition, goal revelation has a positive effect on the aggregate performance of negotiators, and on the likelihood to reach agreement. Further analysis show goal revelation to be a cooperation mechanism by which negotiators are able to iden- tify acceptable agreements in scenarios characterized by few socially (Pareto) beneficial outcomes.

MIT Master's Thesis: Crowdsourcing Moral Psychology (2021)

Completed with the School of Architecture and Planning at the Massachusetts Institute of Technology

https://dspace.mit.edu/handle/1721.1/130838

Ethical trade-off surveys have played a key role in building a data-driven understanding of human moral psychology. They have been conducted all over the world for decades, eliciting assessment of ethical dilemma outcomes from populations as diverse as those of rural, tribal settlements, and industrialized, information-age, cosmopolitan cities. While much data has been gathered through these surveys, attempts to compare what people across cultures consider ethically justifiable have been hindered by the fact that the surveys used have been reformulated for different cultures in the scenarios they depict, and in their framing. The objective of this thesis project is to build a survey tool with global reach and internationalized surveys, in order to collect survey data from around the world using consistent scenarios and framing. Building on the precedent and success of the Moral Machine tool for surveying people around the world regarding ethical dilemmas involving autonomous vehicles, I built and deployed a tool for conducting surveys with scenarios of the classic action/omission trolley problem, to collect ethical dilemma survey data internationally, in ten languages, for three variants of the trolley problem - one for remote action/omission with no double effect consideration, one for double effect consideration with direct action/omission, and one for double effect consideration with remote action/omission. Analyzing data from this experiment, I conclude that differences in preferences across the variants are confirmed across populations, and that they are universal across populations in order of preference.

BUiD Master's Thesis: Empirical Studies in Computer-Mediated Interest-Based Negotiations (2009)

Completed with the Faculty of Informatics at the British University in Dubai (and in association with the School of Informatics of the University of Edinburgh in Scotland, UK)

http://bspace.buid.ac.ae/handle/1234/53

Negotiations in which participants exchange offers based on their chosen positions can be extended to include dialogue about their interests. Revelation of negotiators' interests allows them to make more acceptable o ers and perhaps propose possible alternative approaches toward each other's interests, both of which may result in mutually and individually beneficial outcomes. However, it can also expose their strategies, and possibly their dependencies on other negotiators toward the achievement of their goals. Revealing this information can leave them vulnerable to extortion or retribution, but it can also be used to gain sympathy or build a relationship of trust and reciprocity.This dissertation studies human behaviour and performance upon introducing options for goal inquiry and revelation into mediated-protocol negotiation scenarios. Empirical studies were conducted by having human players negotiate over an alternating o er protocol and an interest-based bargaining protocol, on a platform specially adapted for this purpose. The analysis of data from these experiments revealed interesting patterns in the human use of goal revelation,and its effects on individual and social outcomes and likelihood of agreement. The design of the experiments and the development of the experimentation platform lay the groundwork for the further study of goal revelation in mediated negotiations with humans.

IC2S2 Tutorial: Crowdcomputing and Citizen Science for Large-Scale Experiments

https://www.media.mit.edu/events/ic2s2-tutorial-crowdcomputing-and-citizen-science-for-large-scale-experiments/

Presenters: Snehalkumar ‘Neil’ S. Gaikwad, Sohan Dsouza, Oana Vuculescu, Andrew Mao, Iyad Rahwan
at Maternushaus, Kardinal-Frings-Straße, 1 Cologne, Germany
for The International Conference on Computational Social Science (IC2S2) 2017

Historically, scientific experiments have been conducted at a small scale either with artificial environments or with the expertise of limited number of scientists. While social science literature investigates very deep questions to understand human behavior, many experiments are usually limited by the number of participants and duration of a study. On the contrary, computer science literature exploits advanced computational techniques to crunch voluminous datasets, but research designs are generally not experimental, which limits the opportunity to generate causal inferences.

In this tutorial we demonstrate how crowdcomputing can enable computational social scientists to engage with millions of users on the Internet and study human behavior at scale for a longer time. We showcase pitfalls and lessons learned from various crowdcomputing and citizen science projects.

Furthermore, we provide insights about how to build a sustainable citizen science community to scale science beyond the traditional laboratories. We envisage this tutorial will help computational social scientists effectively use crowdcomputing to investigate deep research questions and longitudinally validate their hypotheses in large scale experiments.