AAAI Symposium on AI for Social Good
In 2016, the CCC co-sponsored a workshop on Artificial Intelligence for Social Good with AAAI and OSTP. In order to further the discussion of the benefits of AI to society, the CCC co-sponsored the AAAI 2017 Spring Symposium on AI for Social Good at Stanford University. This symposium focused on the promise of AI across multiple sectors of society.
Almost any real-world problem, which is important for society’s benefit, and could potentially be solved using AI techniques, is within the ambit of this symposium. For example, in urban computing:
- efficient management of traffic lights allows for smart flow of traffic through a city;
- spatio-temporal GPS data from various sources is used to predict traffic volume at various places and times in cities, which powers recommendation systems providing smart driving directions to commuters;
- GPS trajectories of taxicabs are utilized to detect flawed urban planning of newly built roads and subway lines;
- and aggregate level gas consumption and pollution emissions patterns of vehicles in different parts of a city are predicted using machine learning techniques, which can be used to identify “gas-efficient” driving routes.
In the field of healthcare:
- companion pet bots have been built which provide emotional support and monitor senior citizens for signs of depression;
- AI based virtual nursing assistants have been developed which follow up with patients after they have been discharged from hospitals;
- and sequential planning algorithms have been used to efficiently spread awareness about dangerous diseases such as HIV among disadvantaged populations such as homeless youth.
In the field of public welfare and social justice:
- data-driven early intervention systems have been built which pro-actively identify police officers likely to have adverse interactions with the general public, thereby harming police-public relations;
- decision support systems are in place which identify high-school students who are likely to need additional support so that they can complete high-school in time;
- and game theoretic techniques are used to determine the optimal government policies that would alleviate poverty in the most efficient, sustainable manner.
In the field of sustainability:
- software assistants based on game theory recommend patrol plans to wildlife park rangers for protection of tigers and rhinos (among other animals) from being killed by well-organized poachers;
- spatio-temporal models for bird species distributions are built, which allows accurate visualization of migratory patterns of many birds;
- and sequential planning algorithms are used to decide which areas of coastal habitat to protect in order to minimize sea level rise.
Finally, in the field of security:
- software assistants based on game theory (a subfield of AI) have been developed for generating randomized patrol plans to protect important infrastructure such as ports, airports, flights, transit systems;
- AI techniques are used to decide where to place honeypot bots inside a computer network to fool hackers;
- and multilateral relations between countries are inferred from sparse dyadic political events using machine learning techniques.
While there has been significant progress, there still exist many major challenges facing the design of effective AI based approaches to deal with the difficulties in real-world domains. The symposium served two purposes in this regard. First, the symposium provided an opportunity to showcase real-world deployments of AI based systems for social good. More often than not, unexpected practical challenges emerge when solutions developed in the lab are deployed in the real world, which makes it challenging to utilize complex and well thought out computational/modeling advances. Learning about the challenges faced in these deployments during the symposium will help us understand lessons of moving from the lab to the real world.
There is also a need to build AI based systems which dynamically adapt to changing environments, are robust to errors in execution and planning, and handle uncertainties of different kinds that are common in the real world. Addressing these challenges requires collaboration from different communities including artificial intelligence, game theory, operations research, social science, and psychology. This symposium was structured to encourage a lively exchange of ideas between members from these communities. We encouraged submissions to the symposium from: (i) researchers who have used (or are currently using) AI techniques to solve important real-world problems for society’s benefit in a measurable manner; and (ii) researchers/practitioners/domain experts from new social domains which could benefit from the introduction of AI based systems.
March 27, 2017 (Monday)
|Welcome Talk / Setting the Agenda
|Self-Introductions of Symposium Attendees
|Invited Talk 1 - AI for Sustainability and Public Health
|Invited Talk 2 - AI for the Social Sciences
|Poster Session 1
|Talk Sessions 1: Healthcare
Session Chair: Dr. Eric Horvitz
|Talk Sessions 2: Social Welfare
Session Chair: Dr. Eric Rice
|Poster Session 2
March 28, 2017 (Tuesday)
|Talk Sessions 3: Urban Planning
Session Chair: Dr. Virginia Dignum
|Talk Sessions 4: Computational Sustainability
Session Chair: Dr. Fei Fang
|Talk Sessions 5: Miscellaneous
Session Chair: Amulya Yadav
|Lightning Talks Session 1
|Overarching discussion / Summary of meeting
March 29, 2017 (Wednesday)
|Report Writing / Next Steps
- Eric Horvitz, Microsoft Research
- Barbara Grosz, Harvard University
- Amy Greenwald, Brown University
- David Parkes, Harvard University
- Carla Gomes, Cornell University
- Stephen Smith, Carnegie Mellon University
- Gregory Hager, Johns Hopkins University
- Ann W. Drobnis, Computing Research Association
- Nicole Sintov, Ohio State University
- Milind Tambe, University of Southern California
- Amulya Yadav, University of Southern California
- Fei Fang, Harvard University
- Bryan Wilder, University of Southern California