IC14 – Why Intelligence Tests Are Still a Very Hard Problem for AI

Lecturer: Maithilee Kunda
Fields: Artificial Intelligence/Cognitive Science


While nearly 60 years of AI research on solving intelligence tests has yielded many techniques for many tests, we are still quite far from having an artificial agent that can “sit down and take” an intelligence test without specialized algorithms having been designed for that purpose. This course will discuss: 1) why intelligence tests are such a good challenge for AI; 2) a framework for artificial problem-solving agents with four components: a problem definition, input processing, domain knowledge, and a problem-solving strategy or procedure; 3) several types of agents from my own research that use visual-imagery-based strategies to solve problems from the well-known Raven\’s Progressive Matrices tests; and 4) ways in which an imagery-based agent could learn its domain knowledge, problem-solving strategies, and problem definition/input processing components from experience, instead of each being manually designed. We will also discuss implications of this work in understanding cultural differences in cognition, cultural and racial biases in the history of intelligence tests, and the worrying and still-prevalent phenomenon of stereotype threat.


  • Kunda, M. (2020). AI, visual imagery, and a case study on the challenges posed by human intelligence tests. Proceedings of the National Academy of Sciences, 117(47), 29390-29397. URL: https://doi.org/10.1073/pnas.1912335117
  • Kunda, M. (2019). Nonverbal task learning. In Proceedings of the Seventh Annual Conference on Advances in Cognitive Systems, 609-622. URL: https://cdn.vanderbilt.edu/vu-my/wp-content/uploads/sites/2127/2016/06/30080321/Kunda-2019-Nonverbal-task-learning.pdf
  • Hernández-Orallo, J., Martínez-Plumed, F., Schmid, U., Siebers, M., & Dowe, D. L. (2016). Computer models solving intelligence test problems: Progress and implications. Artificial Intelligence, 230, 74-107. URL: https://doi.org/10.1016/j.artint.2015.09.011
  • Evans, T. G. (1964, April). A heuristic program to solve geometric-analogy problems. In Proceedings of the April 21-23, 1964, spring joint computer conference (pp. 327-338). URL: https://doi.org/10.1145/1464122.1464156


Maithilee Kunda
Dr. Maithilee Kunda

Maithilee Kunda holds a B.S. in mathematics with computer science from MIT and a Ph.D. in computer science from Georgia Tech. She is currently an assistant professor of computer science and computer engineering at Vanderbilt University. Her work in artificial intelligence, in the area of cognitive systems, looks at how visual thinking contributes to learning and intelligent behavior, with a focus on applications for individuals on the autism spectrum. In 2016, she was recognized as a visionary on the MIT Technology Review’s annual list of 35 Innovators Under 35 for her work at the intersection of autism, AI, and visual thinking, and in 2020, her research on visuospatial cognitive assessment was featured on CBS 60 Minutes with Anderson Cooper, as part of a segment on, “Recruiting for talent on the autism spectrum.”

Affiliation: Vanderbilt University
Homepage: https://my.vanderbilt.edu/aivaslab/