How Cognitive Science and Artificial Intelligence Are Intertwined
The I in AI.
Posted June 2, 2022 | Reviewed by Lybi Ma
- If we want to understand the mechanisms behind AI, cognitive science might come to the rescue.
- Artificial intelligence and cognitive science have surprising similarities.
- AI focuses on artificial minds with human minds as an example.
- Cognitive science focuses on human minds with artificial minds as an example.
By now you have likely been persuaded how AI (artificial intelligence) provides all the opportunities one could hope for, and all the threats one should fear. But with all the news about AI that comes to us, it almost seems as if AI is a recent invention that suddenly caught us by surprise and will change society forever. That is an independent discipline coined recently. I am not denying the opportunities AI has offered, as I would not deny the promise any other scientific discipline or method offers. And I am certainly not ignorant with regards to the risks AI poses to us, as I would not ignore any risks any technologies pose to us.
The History of AI
Artificial Intelligence and cognitive science are very much interrelated. AI was first coined in a proposal for a workshop that took place in the summer of 1956 at Dartmouth College. The workshop aimed to find answers to questions such as how to make machines use language, form abstractions and concepts, and problem solve: How can artificial minds be developed that perform similarly to human minds. The meeting was attended by 11 computer scientists, among them Allan Newell and Herb Simon, who won a Nobel Prize for their work two decades later.
A few weeks after the workshop at Dartmouth, a Special Interest Group in Information Theory was held at MIT. That meeting was attended by researchers in psychology, linguistics, computer science, anthropology, neuroscience, and philosophy. At this meeting too questions regarding language, abstractions and concepts, and problem-solving played a central role. These questions were motivated by the central question: How we can better understand the human mind by developing artificial minds. The Special Interest Group meeting was attended by several researchers who also attended the Dartmouth workshop.
The MIT special interest group marked the cognitive revolution and the start of cognitive science. The cognitive revolution can best be characterized by a movement that emphasized the interdisciplinary study of the human mind and its processes, placing special emphasis on similarities between computational processes and cognitive processes, between human minds and artificial minds. It led to what is now known as cognitive science, an interdisciplinary research program comprised of psychology, computer science, neuroscience, linguistics, and related disciplines.
Looking back at those meetings in the 1950s, the birth of artificial intelligence, and the birth of cognitive science, it is almost as if AI is computer science motivated by psychology, and cognitive science psychology motivated by computer science.
Concepts and Methods
The relationship between AI and cognitive science is not restricted to two workshops. Also in the theories, concepts and methods both use there are striking similarities.
Reinforcement learning in AI is obviously derived from reinforcement learning as we know it in psychology. And central to AI nowadays is deep learning, the use of artificial neural networks. These artificial neural networks were inspired by human neural networks. Particularly around the 1980s, these artificial neural networks showed a lot of promise, less so at the time in AI and more so in cognitive science.
The connection between AI and cognitive science can also be seen in the background of leading researchers. Among the researchers who proposed the Dartmouth workshop were John McCarthy, both a computer scientist and cognitive scientist, and Marvin Minsky both a cognitive and computer scientist. Others that attended the workshop, including Allan Newell, had a research background in psychology and computer science. David Rumelhart and Jay McClelland, who led the research in artificial neural networks in the 1980s, both had a background in psychology. And one of the contributors to the two-volume “Parallel Distributed Processing” by Rumelhart and Jay McClelland was Jeff Hinton, seen as one of the leading figures in artificial neural networks, a cognitive psychologist and a computer scientist.
The Cost of Explainability
But there is a more important take-home message for the interdependencies of AI and cognitive science. That take-home message does not lie in the history of AI and cognitive science, neither in the use of similar concepts and methods, nor in the background of the researchers. It lies in what we can learn from AI and cognitive science. For instance, with regards to the importance of Explainable AI also referred to as XAI. Whereas AI (and data science) often focus on accuracy, we may want to pay more attention to why techniques and methods make particular decisions.
The accuracy and performance of AI systems are rapidly increasing with more computing power and more complex algorithms. That also comes at a price: Explainability. If we want to build algorithms that are fair—following the FAIR principles of Findability, Accessibility, Interoperability, and Reuse of digital assets—we need to at least be able to understand the mechanisms behind the algorithms. And that reminds me of what Mike Jones stated a couple of years ago about data science:
Within the Cognitive Sciences, we have been considerably more skeptical of big data’s promise, largely because we place such a high value on explanation over prediction. A core goal of any cognitive scientist is to fully understand the system under investigation, rather than being satisfied with a simple descriptive or predictive theory. (Jones, 2017)
In other words, the interdependencies between AI and cognitive science do not only lie in the past. In fact, more than ever they lie in the present and the future.
Jones, M. N. (2017). Big Data in Cognitive Science. Psychology Press / Taylor & Francis.
Louwerse, M. (2021). Keeping Those Words in Mind: How language creates meaning. Rowman & Littlefield.
Rumelhart, D. E., McClelland, J. L., & PDP Research Group. (1986). Parallel Distributed Processing . MIT Press.