The online interactive magazine of the Association for the Advancement of Artificial Intelligence

Vol 41 No 3: Fall 2020 | Published: 2020-09-24


Challenges of Human-Aware AI Systems

Real-time traffic signal control presents a challenging multiagent planning problem, particularly in urban road networks where, unlike simpler arterial settings, there are competing dominant traffic flows that shift through the day. Further complicating matters, urban environments require attention to multimodal traffic flows (vehicles, pedestrians, bicyclists, buses) that move at different speeds and may be given different priorities. For the past several years, my research group has been developing and refining a real-time, adaptive traffic signal control system to address these challenges, referred to as scalable urban traffic control (Surtrac). Combining principles from automated planning and scheduling, multiagent systems, and traffic theory, Surtrac treats traffic signal control as a decentralized online planning process. In operation, each intersection repeatedly generates and executes (in rolling horizon fashion) signal-timing plans that optimize the movement of currently sensed approaching traffic through the intersection. Each time a new plan is produced (nominally every couple of seconds), the intersection communicates to its downstream neighbors what traffic it expects to send their way, allowing intersections to construct longer horizon plans and achieve coordinated behavior. Initial evaluation of Surtrac in the field has demonstrated significant performance improvements, and the technology is now deployed and operating in several U.S. cities. More recent work has focused on integrating real-time adaptive signal control with emerging connected vehicle technology, and exploration of the opportunities for enhanced mobility that direct vehicle (or pedestrian) to infrastructure communication can provide. Current technology development efforts center on vehicle route sharing, smart transit priority, safe intersection crossing for pedestrians with disabilities, real-time incident detection, and integrated optimization of signal control and route choice decisions. This article provides an overview of this overall research effort.


Stephen F. Smith is a research professor at the Robotics Institute, Carnegie Mellon University, where he heads the Intelligent Coordination and Logistics Laboratory. He is also chief scientist of Rapid Flow Technologies Inc., a company he cofounded to commercialize the Surtrac technology. Smith’s research focuses broadly on the theory and practice of next-generation technologies for planning, scheduling, and coordination. His current research interests include execution-driven planning and scheduling systems, multirobot path fi nding, agent-based models for distributed task and resource allocation, and intelligent transportation systems technologies.


Conversational Intelligence Challenge: Accelerating Research with Crowd Science and Open Source

Sketching is a valuable but underutilized tool for science education. Sketch worksheets were developed to help change this, by using artificial intelligence technology to give students immediate feedback and to give instructors assistance in grading. Sketch worksheets use automatically computed visual representations combined with conceptual information to give feedback to students, by computing analogies between students’ sketches and an instructor’s solution sketch. This enables domain experts to develop sketch worksheets, to facilitate dissemination. We describe our experiences in deploying them in geoscience and artificial intelligence classes. The geoscience worksheets, authored by geoscientists at University of Wisconsin– Madison, were used at both Wisconsin and Northwestern University. The artificial intelligence worksheets were developed and used at Northwestern. Our experience indicates that sketch worksheets can provide helpful on-the-spot feedback to students, and significantly improve grading efficiency, to the point where sketching assignments can be more practical to use broadly in science, technology, engineering, and mathematics education.


Kenneth D. Forbus is the Walter P. Murphy Professor of Computer Science and Professor of Education at Northwestern University. He received his degrees from the Massachusetts Institute of Technology (PhD in 1984). His research interests include qualitative reasoning, analogical reasoning and learning, spatial reasoning, sketch understanding, natural language understanding, cognitive architecture, reasoning system design, intelligent educational software, and the use of AI in interactive entertainment. He is a Fellow of the Association for the Advancement of Artifi cial Intelligence, the Cognitive Science Society, and the Association for Computing Machinery. He is the inaugural recipient of the Herbert A. Simon Prize, a recipient of the Humboldt Research Award, and has served as Chair of the Cognitive Science Society.

Bridget Garnier is a PhD candidate in the Department of Geosciences at the University of Wisconsin–Madison. Basil Tikoff is a geologist and professor at the University of Wisconsin–Madison. In addition to disciplinary studies in the geological sciences, he has worked with cognitive scientists and geoscience educators on spatial learning within the natural sciences.

Wayne Marko is an adjunct earth science instructor at several community colleges in the Chicago area. His research interests are in the areas of igneous petrology, magma emplacement, and tectonic environments.

Madeline Usher is a computer programmer and researcher at Northwestern University. Her research interests include spatial and qualitative reasoning. She is the chief programmer for CogSketch, an open-domain sketch understanding and spatial reasoning system.

Matthew McLure completed his PhD in computer science from Northwestern University in March 2019. His research focuses on machine learning via analogical matching and structured spatial representations.


Adaptable Conversational Machines

In complete contract theory, the main approach to limit moral hazard is through modifying incentives for the agents. However, such modifications are not always feasible. One prominent example is Chinese government service provision. Over the years, it has been plagued with inefficiencies as a result of moral hazard. Previous attempts to address these challenges are not effective, as reforms on civil servant incentives face stiff hindrance. In this article, we report an alternative platform — SmartHS — to address these challenges in China without modifying incentives. Through dynamic teamwork, automation of key steps involved in service provision, and improved transparency with the help of artificial intelligence, it places civil servants into an environment that promotes efficiency and reduces the opportunities for moral hazard. Deployment tests in the field of social insurance service provision in three Chinese cities involving close to 3 million social insurance service cases per year demonstrated that the proposed approach significantly reduces moral hazard symptoms. The findings are useful for informing current policy discussions on government reform in China and have the potential to address long-standing problems in government service provision to benefit almost one fifth of the world’s population.


Yongqing Zheng is the chief executive officer and deputy chairman of Shanda Dareway Software Pte Ltd. He is also a professor at the School of Software Engineering, Shandong University, China. His research focuses on AI and the application of this technology to solve real-world problems.

Han Yu is a Nanyang assistant professor at the School of Computer Science and Engineering, Nanyang Technological University, Singapore. He obtained his PhD from the School of Computer Science and Engineering, Nanyang Technological University. His research focuses on real-time, AI-powered, data-driven algorithmic management for complex collaborative networks. He has published over 100 research papers in the form of book chapters; proceedings for leading international conferences such as the Association for the Advancement of Artificial Intelligence, the International Joint Conferences on Artificial Intelligence, and the International Conference on Autonomous Agents and Multiagent Systems; and as articles in journals including AI Magazine, Modern Machinery Science Journal, ACM/ IEEE Transactions, and Scientific Reports. His research work has won 11 conference and journal awards.

Lizhen Cui is a professor and the vice chair of the School of Software Engineering, Shandong University, China. Between 2013 and 2014, he was a visiting scholar at the Georgia Institute of Technology in the United States. His main research interest includes data science and engineering, intelligent data analysis, service computing, and collaborative computing.

Chunyan Miao is a chair professor at the School of Computer Science and Engineering, Nanyang Technological University, Singapore. She received her BSc degree from Shandong University, and her MEng and PhD from Nanyang Technological University. She is the Founding Director of the LILY Research Centre and the Alibaba–Nanyang Technological University Joint Research Institute. Her research focuses on humanized AI.

Cyril Leung is a professor at the Department of Electrical and Computer Engineering at the University of British Columbia, Vancouver, Canada. He received his BSc (first class honors) degree from Imperial College, University of London, England, and his MS and PhD degrees in electrical engineering from Stanford University. His research interests include digital communications, wireless communication networks, sensor networks, ubiquitous computing, elderly-friendly technologies, digital/social signal processing, security and privacy, trust computational models, and information theory.

Yang Liu is a senior researcher in the AI Department of WeBank, China. Her research interests include machine learning, federated learning, transfer learning, multi-agent systems, statistical mechanics, and applications of these technologies in the financial industry. She received her PhD from Princeton University in 2012 and her Bachelor’s degree from Tsinghua University in 2007. She holds multiple patents. Her research has been published in leading scientific journals such as ACM Transactions on Intelligent Systems and Technology and Nature.

Qiang Yang is a chair professor at the Department of Computer Science and Engineering of the Hong Kong University of Science and Technology. He received his MSc degree in Astrophysics, MSc in Computer Science, and PhD in Computer Science at the University of Maryland in the US. His research interests include transfer learning, machine learning, planning, and data mining in AI. He is also the president of the International Joint Conferences on Artificial Intelligence, an executive council member of the Association for the Advancement of Artificial Intelligence, and editor-in-chief of IEEE Transactions on Big Data. He was the founding editor-in-chief of ACM Transactions on Intelligent Systems and Technology. He was the founding director of the Hong Kong University of Science and Technology’s Big Data Institute. He is a fellow of the Association for Computing Machinery, the Institute of Electrical and Electronics Engineers, the Association for the Advancement of Artificial Intelligence, the American Association for the Advancement of Science, the International Association for Pattern Recognition, and the Chinese Association for Artificial Intelligence.


Automated Assignment of Helpdesk Email Tickets: An AI Lifecycle Case Study

Conversion rate optimization (CRO) means designing an e-commerce web interface so that as many users as possible take a desired action such as registering for an account, requesting a contact, or making a purchase. Such design is usually done by hand, evaluating one change at a time through A/B testing, evaluating all combinations of two or three variables through multivariate testing, or evaluating multiple variables independently. Traditional CRO is thus limited to a small fraction of the design space only, and often misses important interactions between the design variables. This article describes Ascend by Evolv, 1 an automatic CRO system that uses evolutionary search to discover effective web interfaces given a human-designed search space. Design candidates are evaluated in parallel online with real users, making it possible to discover and use interactions between the design elements that are difficult to identify otherwise. A commercial product since September 2016, Ascend has been applied to numerous web interfaces across industries and search space sizes, with up to fourfold improvements over human design. Ascend can therefore be seen as massively multivariate CRO made possible by artificial intelligence.


Risto Miikkulainen is a Professor of Computer Science at the University of Texas at Austin and Associate Vice President of evolutionary AI at Cognizant. His current research focuses on methods and applications of neuroevolution, as well as neural network models of natural language processing and vision. At Cognizant, and previously as Chief Technology Officer of Sentient Technologies, he is scaling up these approaches to real-world problems. He received an MS in Engineering from the Helsinki University of Technology (now Aalto University) in 1986, and a PhD in Computer Science from the University of California-Los Angeles in 1990.

Jonathan Epstein is Chief Strategy Officer for Evolv Technologies. Previously Chief Marketing Officer and Senior Vice President (International) at Sentient, he was intimately involved with the development and launch of Ascend. Before working at Evolv and Sentient, Epstein has held key executive positions at the intersection of technology and media, including President of Omek Interactive, Chief Executive Officer of GameSpot, Chief Executive Officer of Double Fusion, and Senior Vice President of IGN Entertainment. He has authored multiple patents in fields ranging from gesture control to in-game advertising to remotely operated underwater vehicles. Epstein graduated from Harvard University with an AB in Physical Sciences.

Babak Hodjat is Vice President of Evolutionary AI at Cognizant, and former cofounder and Chief Executive Officer of Sentient and a cofounder of Sentient Investment Management. He is a serial entrepreneur, having started a number of Silicon Valley companies as main inventor and technologist. Before cofounding Sentient, Hodjat was Senior Director of Engineering at Sybase iAnywhere, where he led mobile solutions engineering, and a cofounder, Chief Technology Officer, and board member of Dejima Inc. Hodjat is the primary inventor of Dejima’s agent-oriented technology applied to intelligent interfaces for mobile and enterprise computing — the technology behind Apple’s Siri. He has publications and patents in numerous fields of AI, including natural language processing, machine learning, genetic algorithms, and distributed AI. He holds a PhD in Machine Intelligence from Kyushu University, Fukuoka, Japan.

Neil Iscoe was the Chief Executive Officer and cofounder of Digital Certainty, the company that created the original version of Ascend. After the product was sold to Sentient Technologies, he became the product’s general manager. Previously, he was the Associate Vice President of Research and Director of Technology Commercialization at the University of Texas, where he was responsible for creating commercialization entities and marketable products from university research. In 2011, he left the university to build Ascend. He has an MS and PhD in Computer Science, with a Specialization in Systems and AI from the University of Texas.

Jingbo Jiang was a intern in Sentient Technologies during the summer of 2018, working on an evolution algorithm and its application to web page design, and in particular the comparisons with the Taguchi method. Her background is in machine learning, computer vision, and language processing. She earned her MS in Data Science from the University of Pennsylvania in 2019 and her BE in Electrical Engineering from Beihang University, China in 2017.

Diego Legrand was a Machine Learning Engineer at Sentient Technologies. His work on Bayesian models changed Ascend’s way of finding top performing candidates and significantly increased the learning speed of the system. Diego has a masters in Applied Mathematics from Centrale University in Paris, France.

Sam Nazari is the Vice President of Customer Success at Evolv.AI. He leads the global team that works closely with clients to help them understand and integrate AI across their enterprises (from large Fortune-500 companies to medium-sized businesses spanning multiple verticals). This ranges from explaining best-use cases for AI in marketing through to how to prepare their data, design and implement the technology, and seek compliance with their Information Technology teams, to the successful rollout of AI to help drive revenue. Nazari has a BS in Computer and Electrical Engineering from the University of Utah.

Xin Qiu is a senior research scientist at Cognizant and previously at Sentient. His research interests include evolutionary computation, probabilistic modeling, bandit algorithms, and reinforcement learning. He earned his PhD from the National University of Singapore in 2016 and his BE from Nanjing University in 2012.

Michael Scharff is the chief executive officer for Evolv Technologies, the AI fi rm behind the Ascend autonomous website optimization platform. Scharff brings over two decades of digital commerce and retail experience, with leadership roles at some of the most well-known retailers in North America including Sears Canada, Toys R Us, Staples, and Best Buy. He has a wealth of experience in all aspects of retailing and across numerous industry verticals and channels. Scharff has built and managed highly successful omni-channel and global eCommerce businesses, and led teams in merchandising, digital marketing, innovation, and other functional areas.

Cory Schoolland has over a decade of experience heading marketing design efforts for San Francisco-based softwareas-a-service companies including RichRelevance, Sentient Technologies, and Evolv Technologies. He believes in the power of good design to improve our lives, and enjoys combining words, shapes, images, and colors to tell a story, or taking existing content and making it beautiful.

Rob Severn is the director of product at the Evolv. He is responsible for understanding the optimization market’s problems and needs to better guide Evolv’s optimization product. Severn received a BA in Mathematics from Cambridge University in 2006.

Aaron Shagrin has been working with technology companies, large and small, for over 20 years. He has been a part of multiple startups, Fortune-500 companies, and private equity firms. He has a deep background in product management and strategy, acquisitions, alliances, and business development and sales. He has a BBA from The University of Texas.


Large Scale Personalized Categorization of Financial Transactions

Artificial intelligence, whether embodied as robots or Internet of Things, or disembodied as intelligent agents or decision-support systems, can enrich the human experience. It will also fail and cause harms, including physical injury and financial loss as well as more subtle harms such as instantiating human bias or undermining individual dignity. These failures could have a disproportionate impact because strange, new, and unpredictable dangers may lead to public discomfort and rejection of artificial intelligence. Two possible approaches to mitigating these risks are the hard power of regulating artificial intelligence, to ensure it is safe, and the soft power of risk communication, which engages the public and builds trust. These approaches are complementary and both should be implemented as artificial intelligence becomes increasingly prevalent in daily life.


Aaron Mannes is a senior policy advisor with Culmen International, LLC, where he supports the Department of Homeland Security Science & Technology Directorate’s Data Analytics Technology Center. Mannes has previously worked at the University of Maryland Institute for Advanced Computer Studies modeling terrorism and international security affairs. He earned his doctorate in policy studies from the University of Maryland College Park and has written several books and numerous articles on international affairs and technology policy.


Improving the Accuracy and Transparency of Underwriting with AI to Transform the Life Insurance Industry

The factors that define and influence the success of industry–academic research in artificial intelligence have evolved significantly in the last decade. In this article, we consider what success means from both sides of a collaboration and offer our perspectives on how to approach the opportunities and challenges that come with achieving success. These perspectives are grounded on the recent and significant investments that have been made between IBM and several higher education institutions around the world, including IBM’s Artificial Intelligence Horizons Network, the Massachusetts Institute of Technology–IBM Watson Artificial Intelligence Lab, and the Massachusetts Institute of Technology Quest for Intelligence.


Lisa Amini is the director of IBM Research Cambridge, which is home to the MIT–IBM Watson AI Lab and IBM Research’s AIHN. She was previously director of knowledge and reasoning research in the Cognitive Computing group at IBM’s TJ Watson Research Center in New York, and she is also an IBM Distinguished Engineer. Amini was the founding director of IBM Research Ireland, and the first woman laboratory director of any non-US IBM Research Global Lab (2010–2013). As senior manager of the Exploratory Stream Processing Research Group, Amini was the founding chief architect for IBM’s InfoSphere Streams product and research predecessor. She earned her PhD in computer science from Columbia University in New York.

Ching-Hua Chen is a research staff member and manager at IBM Research. Her current research involves developing computational methods to enable technologies for health behavior modification. She is coprincipal investigator of the Center for Health Empowerment, Analytics, Learning and Semantics, which is a collaboration between IBM and Rensselaer Polytechnic Institute within the AIHN. Previously, she was the research director of the IBM Research Collaboratory in Singapore, where she collaborated with the Singapore government on solutions for traffic management. She earned her PhD in operations research and business administration from the Pennsylvania State University.

David Cox is the IBM director of the MIT–IBM Watson AI Lab, a first of its kind industry–academic collaboration between IBM and MIT, focused on fundamental research in AI. Prior to joining IBM, Cox was the John L. Loeb Associate Professor of the Natural Sciences and of Engineering and Applied Sciences at Harvard University, where he held appointments in computer science, the Department of Molecular and Cellular Biology and the Center for Brain Science. Cox’s ongoing research is primarily focused on bringing insights from neuroscience into machine learning and computer vision research. He is a faculty associate at the Berkman–Klein Center for Internet and Society at Harvard Law School and is an agenda contributor at the World Economic Forum. He has received a variety of honors, including the Richard and Susan Smith Foundation award for excellence in biomedical research, the Google Faculty Research award in computer science, and the Roslyn Abramson award for excellence in undergraduate teaching. His academic laboratory has spawned several startups across a range of industries, ranging from AI for healthcare to autonomous vehicles.

Aude Oliva is the executive director of the MIT–IBM Watson AI Lab and The MIT Quest for Intelligence. She is also a principal research scientist at the Computer Science and Artificial Intelligence Laboratory. She formerly served as an expert to the National Science Foundation, Directorate of Computer and Information Science and Engineering. Her research interests span computer vision, cognitive science, and human neuroscience. She was honored with the National Science Foundation CAREER Award, a Guggenheim Fellowship, and the Vannevar Bush Faculty Fellowship. She earned a MS and PhD in cognitive science from the Institut National Polytechnique de Grenoble, France.

Antonio Torralba is a professor of electrical engineering and computer science at MIT and a principal investigator at the Computer Science and Artificial Intelligence Laboratory. He is also the MIT director of the MIT–IBM Watson AI Lab and the inaugural director of The MIT Quest for Intelligence. Torralba researches computer vision, machine learning, and human visual perception, with an interest in building systems that can perceive the world the way humans do. He was honored with a National Science Foundation CAREER award, the J.K. Aggarwal Prize from the International Association for Pattern Recognition, the Frank Quick Faculty Research Innovation Fellowship, and the Louis D. Smullin (’39) Award for Teaching Excellence. Torralba earned a BS from TelecomBCN, and a PhD from the Institut National Polytechnique de Grenoble, France. He did his postdoctoral work at MIT.


Scaling Up Data-Driven Pilot Projects

This article discusses a set of principles for accountability and transparency in AI as well as a set of antipatterns or harmful trends too often seen in deployed systems. It provides concrete suggestions for what can be done to shift the balance away from these antipatterns and toward more positive ones.


Jeanna Neefe Matthews is an associate professor of computer science at Clarkson University. She is a member of the ACM’s Technology Policy Council and the chair of the AI and Algorithmic Accountability subcommittee. She is an affiliate at Data and Society in Manhattan.


The Reproducibility Crisis Is Real

Begin with the end in mind! 1 PhD students in artificial intelligence can start to prepare for their career after their PhD degree immediately when joining graduate school, and probably in many more ways than they think. To help them with that, I asked current PhD students and recent PhD computer science graduates from the University of Southern California and my own PhD students to recount the important lessons they learned (perhaps too late) and added the advice of Nobel Prize and Turing Award winners and many other researchers (including my own reflections), to create this article.


Sven Koenig is a professor in computer science at the University of Southern California. Most of his research centers around techniques for decision-making (planning and learning) that enable single situated agents (such as robots or decision-support systems) and teams of agents to act intelligently in their environments and exhibit goal-directed behavior in real-time, even if they have only incomplete knowledge of their environments, imperfect abilities to manipulate them, limited or noisy perception, or insufficient reasoning speeds. Additional information can be found on his webpages at .


Addendum - Special Thanks to this issue's guest editors

We would like to acknowledge the following guest editors for their contributions to this issue’s Conversational AI Special Track articles: 

  • Alborz Geramifard, Facebook AI
  • Dilek Hakkani-Tur, Amazon Alexa AI
  • Peter Henderson, Stanford University
  • Alex Rudnicky, Carnegie Mellon University
  • Asli Celikyilmaz, Microsoft Research

Thank you for your continued invested time and effort that has made this Conversational AI Special Track a success.