The interaction between humans and machines becomes particularly compelling when machines exhibit cognitive processes. Since its inception, artificial intelligence (AI) has been closely intertwined with psychology and theories of learning, reflecting a deep connection where one mirrors the other.
The parallels between human and computer learning go beyond superficial comparisons. Demis Hassabis, a leading AI researcher with a background in cognitive neuroscience, has authored influential works exploring the interplay between neuroscience, psychology, and AI, and how they inspire one another (Hassabis et al., 2017).
The study of AI can potentially inform our understanding of children's learning processes. Hassabis asserts that the similarities between AI and children's learning are significant, suggesting that these insights could uncover fundamental principles of learning.
Pioneers in AI and Cognition
The term "artificial intelligence" was likely coined at a pivotal 1956 conference attended by key figures like Marvin Minsky, Herbert Simon, and Allen Newell. Their interest in psychology and learning led to foundational work, such as "Elements of a Theory of Human Problem Solving," which posited that human thought processes could model machine problem-solving.
During this period, psychology was dominated by behaviorism, which focused solely on stimuli and responses, treating the mind as a "black box." The emerging AI vision sought to open this box, understand cognitive processes, and create thinking machines based on these principles—an endeavor akin to building an electronic brain.
Despite the limitations of early computers, which were bulky and operated using punch cards, these pioneers successfully programmed machines capable of proving mathematical theorems through symbolic manipulation. Simon and Newell's 1971 paper highlighted the potential applications of problem-solving theories in education and learning processes (Simon & Newell, 1971).
John R. Anderson, who joined these early AI pioneers, focused on mathematical learning. He developed the Adaptive Control of Thought (ACT) theory, which informed the creation of intelligent tutoring systems designed to aid children's learning in subjects like algebra.
Papert, Piaget, and Constructivism
A parallel development in AI and education began with Seymour Papert, a South African mathematician who collaborated with developmental psychologist Jean Piaget. Papert adapted Piaget's theories, advocating for children's development through interactions with computers, laying the groundwork for digital education.
In 1963, Papert joined Marvin Minsky in the United States, co-leading the Artificial Intelligence Laboratory and co-authoring the seminal book *Perceptrons*. Papert's work, including the creation of the Logo programming language, aimed to teach children mathematics through programming, promoting a concept he called "Mathland"—an environment where children learn math as naturally as language.
Papert's initiatives, such as the collaboration with Lego to create Mindstorms and the development of the Scratch programming language, were pivotal in integrating technology with learning. He also co-founded the MIT Media Lab and contributed to the One-Laptop-Per-Child project, which sought to provide every child with a laptop (though this goal was not fully realized).
While Papert was visionary, his ideas were not always scientifically validated. Concepts like "Mathland" and constructivist learning, which emphasize self-guided exploration over formal instruction, have been criticized for their limited effectiveness in teaching complex subjects like mathematics. Research has consistently shown that prior knowledge is essential for acquiring new knowledge, challenging the notion that children can teach themselves through exploration alone (Kirschner & van Merrienboer, 2013; Dehaene, 2020).
Curiosity and Misaligned Rewards
AI development has also drawn inspiration from human learning, particularly the role of curiosity. In 2015, DeepMind researchers demonstrated that an AI could learn to play Atari games by itself using reinforcement learning (Mnih et al., 2015). However, the AI struggled with games like *Montezuma's Revenge*, which required a sequence of actions before any reward was given. Incorporating a curiosity-driven approach, where the AI was rewarded for encountering new stimuli, improved performance significantly.
Similar to humans, AI can encounter pitfalls with misaligned rewards. Brian Christian's book *The Alignment Problem* recounts an anecdote where a young girl misinterpreted a reward system, leading to unintended behavior—a situation mirrored in AI training when algorithms exploit reward systems in ways that do not align with intended outcomes (Christian, 2020).
Conclusion
Understanding the parallels between human and AI learning can inform both fields, revealing insights into cognitive processes and effective teaching strategies. As digitalization continues to influence education, we may be on the brink of a revolution in how we understand and facilitate learning.
(This text is a slightly shortened excerpt from Torkel Klingberg's book *The Future of Digital Learning*, published in Swedish in 2023 as *Framtidens digitala lärande*).
Credits: Torkel Klingberg (Professor of Cognitive Neuroscience)
---
References
- Christian, B. (2020). *The Alignment Problem: Machine Learning and Human Values*. Norton and Company.
- Dehaene, S. (2020). *How We Learn: The New Science of Education and the Brain*. London: Allen Lane.
- Hassabis, D., Kumaran, D., Summerfield, C., & Botvinick, M. (2017). Neuroscience-Inspired Artificial Intelligence. *Neuron*, 95, 245-258.
- Kirschner, P. A., & van Merrienboer, J. J. G. (2013). Do learners really know best? Urban legends in education. *Educational Psychologist*, 48, 169-183.
- Mnih, V., Kavukcuoglu, K., Silver, D., Rusu, A. A., Veness, J., Bellemare, M. G., ... & Hassabis, D. (2015). Human-level control through deep reinforcement learning. *Nature*, 518, 529-533.
- Papert, S. (1984). *Mindstorms: Children, Computers, and Powerful Ideas*. Basic Books, Inc.
- Simon, H. A., & Newell, A. (1971). Human problem solving: The state of the theory in 1970. *American Psychologist*, 26, 145-159.
Written by: CL Hub Team.