Artificial cognition: How experimental psychology can help generate explainable artificial intelligence

Abstract

Artificial intelligence powered by deep neural networks has reached a level of complexity where it can be difficult or impossible to express how a model makes its decisions. This black-box problem is especially concerning when the model makes decisions with consequences for human well-being. In response, an emerging field called explainable artificial intelligence (XAI) aims to increase the interpretability, fairness, and transparency of machine learning. In this paper, we describe how cognitive psychologists can make contributions to XAI. The human mind is also a black box, and cognitive psychologists have over one hundred and fifty years of experience modeling it through experimentation. We ought to translate the methods and rigour of cognitive psychology to the study of artificial black boxes in the service of explainability. We provide a review of XAI for psychologists, arguing that current methods possess a blind spot that can be complemented by the experimental cognitive tradition. We also provide a framework for research in XAI, highlight exemplary cases of experimentation within XAI inspired by psychological science, and provide a tutorial on experimenting with machines. We end by noting the advantages of an experimental approach and invite other psychologists to conduct research in this exciting new field.

Explainable artificial intelligence: an analytical review

“…With the above concerns, XAI can help bridge the gap between DL and neuroscience in a mutually beneficial way. On one side, neuroscience and psychology can help build rationalized XAI models that are more easily understood by humankind (Byrne, 2019;Taylor & Taylor, 2020). On the other side, XAI models derived from deep neural networks can also help in understanding the mechanisms of intelligence in the human brain (Fellous et al, 2020;VU et al, 2018).…”

Overview of Explainable Artificial Intelligence for Prognostic and Health Management of Industrial Assets Based on Preferred Reporting Items for Systematic Reviews and Meta-Analyses

“…Though very powerful, many AI methods are black boxes in nature, meaning that the inner mechanism to produce outputs in these methods are unknown [ 28 , 29 ]. Obviously, this opacity is an obstacle in AI penetration across many sensitive or high-stake areas such as banking, defense, finance, and medical areas, even in the common industry [ 30 , 31 ].…”

“…More recently, as the discipline grows, more specialized works emerged. Reviews on XAI have been related to drug discovery [ 31 ], fintech management [ 35 ], healthcare [ 30 , 33 , 36 ], neurorobotics [ 39 ], pathology [ 28 ], plant biology [ 37 ], and psychology [ 29 ]. Thus, it is necessary to produce an analytical compilation of PHM-XAI works, which is still absent.…”

User Stress in Artificial Intelligence: Modeling in Case of System Failure

“…Previous studies have demonstrated several standard approaches to assessing human emotional states and cognitive processes. Research [24] discussed the prospect of using different approaches to evaluate cognitive processes in AI, including machine learning. They described the possibility of using machine learning to increase the efficiency of explainable AI in decision-making for the wellbeing of people.…”

“…Researchers and designers have long recognized the importance of modeling stress and trust as significant influences on the acceptance and adoption of new technologies. On the basis of the aforementioned studies [6], [7], [8], [16], [24], [25], [26], the standard approaches to evaluate cognitive processes and human emotional states can be divided into the following five main groups: 1) survey to measure qualitative characteristics of an AI system; 2) regression modeling; 3) exploratory and confirmatory factor analysis including TAM; 4) predictive modeling; and 5) advanced machine learning modeling (such as random forests and support vector machines). In many studies, including the present research, user stress based on trust in the AI system depends on the reliability of the system and the success of the task performed.…”