Educating a synthetic student

Mike Mozer, University of Colorado Boulder and Google Brain

Presentation link

In educational data mining, we use behavioral data to better understand learners and the nature of learning. I hope you agree with this characterization because it says nothing about whether the learner is a biological or a synthetic agent. In this talk, I focus on deep neural networks as learners. Modern nets have become so complex that we need to use experimental and modeling tools to characterize their behavior, just as we do for humans. Such analyses support the design of robust AI agents, but they also help to identify human-surrogate models that can be used to optimize instructional methods for people.

Bio: Michael Mozer received a Ph.D. in Cognitive Science at the University of California at San Diego in 1987. Following a postdoctoral fellowship with Geoffrey Hinton at the University of Toronto, he joined the faculty at the University of Colorado at Boulder and is presently a Professor in the Department of Computer Science and the Institute of Cognitive Science, as well as a Visiting Faculty Researcher at Google Brain (Mountain View). He is secretary of the Neural Information Processing Systems (NeurIPS) Foundation, has served as Program Chair and General Chair at NeurIPS and as chair of the Cognitive Science Society. He is interested in human-centric artificial intelligence, which involves designing machine learning methods that leverage insights from human cognition, and building software tools to optimize human performance using machine learning methods.

Reconsidering two sigma: Educational data mining for the complete instructional system

Steve Ritter, Carnegie Learning

Presentation link

Educational software developers have long focused on achieving Bloom’s (1984) goal of producing software that is as effective as a personal human tutor. Much progress has been made and, by some measures, this goal has been achieved (vanLehn, 2011). But maybe this is the wrong goal. The “two sigma” problem sets up a competition between teacher-based instruction and educational software. In the classrooms where Carnegie Learning’s MATHia software is most used, there is no need to pick a winner. Most student usage of the software is in the presence of a teacher, and students can benefit from both. A better goal than improving on the teacher’s instruction, then, may be to think of the teacher and the software as cooperating instructional resources with differing characteristics.

This perspective on the goal of educational software can lead to better design decisions focused on improving the overall educational experience for the student. Instead of attempting to reduce the teacher’s classroom role, we focus on focusing the teacher on opportunities to have the biggest impact. During class, the software and the teacher communicate insights about student abilities and needs, enabling each to do a better job. In this talk I will illustrate ways that we use data to provide teachers with insights that help them assist students using MATHia. The teacher, in turn, provides MATHia with information that enables the software to perform better. While this is a new perspective for us, we are starting to see how this systems approach has led to even stronger educational outcomes for students.

Bio: Steve Ritter is Founder and Chief Scientist at Carnegie Learning. He has been developing and evaluating educational systems for over 20 years. He earned his Ph.D. in Cognitive Psychology at Carnegie Mellon University and helped start Carnegie Learning to ensure that the successful Cognitive Tutor approach was available to students across the country. Carnegie Learning’s blended mathematics curricula, including MATHia software (formerly called Cognitive Tutor) are currently used by over 500,000 students each year.
Dr. Ritter was instrumental in the development and evaluation of the Cognitive Tutors for mathematics. He is the author of numerous papers on the design, architecture and evaluation of Intelligent Tutoring Systems and other advanced educational technology and is recognized as an expert on the design and evaluation of educational technology and on educational analytics. He is lead author of an evaluation that is one of the few to be judged by the US Department of Education’s What Works Clearinghouse as meeting their standards without reservations. He have received several awards, including the Best Paper award at the International Conference on Educational Data Mining.
Dr. Ritter leads Carnegie Learning’s research group, which is charged with focusing on improving student outcomes through data mining and field experimentation.

On the Ethics of Data-Powered Instruction, Recommendation, Persuasion, and Behaviour Change

Julita Vassileva, University of Saskatchewan

Presentation link

Increasingly, ethical issues, such as Fairness, Transparency, Accountability, have attracted the attention of researchers in applied AI and Machine Learning, as well as the general public. This talk presents an overview of ethical issues related to using data mining for personalisation in educational and behaviour change applications. It proposes a new framework to discuss the ethical issues in such applications and demonstrates its application with examples of instructional planning and recommender systems.

Bio: Dr. Julita Vassileva is a professor at the University of Saskatchewan and a leading researcher in user modelling, personalization and social computing, as well as AI in Education. She has authored over 250 research papers in the areas of intelligent tutoring and recommender systems, user modeling and personalization, trust and reputation, persuasion and behavior change. She is a member of the editorial boards of several reputed journals, including User Modeling and User Adapted Interaction, IEEE Transactions of Learning Technologies, ACM Transactions on Social Computing, International Journal of AI in Education and is the founding Editor-in-Chief of Frontiers in AI section on AI Supporting Human Learning and Behaviour Change. Dr. Vassileva was elected in 2017 to the Board of CS-Can |Info-Can (the Canadian Computer Science Society), and serves as a Chair of the Research Committee of the Board. Over her 20 years on faculty, she has advised over 35 graduate students (10 PhD and 27 MSc) and was nominated by her students and received the two main awards of the University of Saskatchewan for graduate supervision: the University’s Distinguished Graduate Supervisor Award (2014), and the Advising Excellence Award by the Graduate Student Association (2013). Dr. Vassileva held the NSERC/Cameco Chair for Women in Science and Engineering for the Prairies region (2005 – 2011) and started a science outreach program for aboriginal youth in the North, called “Science Ambassadors”, which has now completed 12 years of impactful activity in over 25 communities and has engaged over 25,000 indigenous youth in science. For her leadership and mentorship she was awarded the Saskatoon’s YWCA “Women of Distinction Award (Science and Research)” in 2015.