Chapter 21 - The Psychology of Artificial Intelligence

Chapter Summary

Key terms

• There are various definitions of artificial intelligence (AI), with one of the most famous being ‘the science and engineering of making intelligent machines, especially intelligent computer systems’ (McCarthy, 2007, p. 2).
• Other key terms in the area include robotics, chatbots and Non-Player Characters (NPCs).
• When individuals discount an AI technology as not really being intelligent, it is termed the AI effect.

Defining and testing artificial intelligence

• It is difficult to test for and define intelligence.
• AIs in science fiction often portray high intelligence across many of Gardner’s (1983) multiple intelligences.
• The Turing test was proposed in 1950 as a measure of testing AI. It requires a computer to have a conversation with a human without the human knowing that they are talking to a machine.
• However, there are critics of the Turing test as an appropriate measure of artificial intelligence, such as Searle’s Chinese room (1980).
• Weak AI occurs where the computer does not actually have intelligence, but only a simulation of it.
• Strong AI occurs where the computer possesses the intelligence that it appears to have.

Social robotics

• Social robotics examines how humans interact socially with robots and how robots interact socially with each other.
• Much research in this field has examined human interaction with high-end commercially available toys that simulate pets.
• Frequently this research has examined robot interactions with children.
• A robot’s appearance and performance affect our engagement with them.
• It is important that we are able to accurately interpret the robot’s actions.
• We appear to prefer robots with similar personality types to ourselves, at least with regard to introversion/extroversion.
• Authors in this field have identified the importance of the robot understanding what the human intends to do and responding to these intentions appropriately.
• Research in social robotics sometimes uses a design called a ‘Wizard of Oz’ study, where the robot is actually controlled by a human instead of AI. There are some ethical implications of this type of research.

Applications of artificial intelligence

• While there are many applications of artificial intelligence, the key psychological ones include gaming, customer service/user interfaces, psychotherapy, companionship and cognitive science.

Cognitive psychology

• Cognitive psychologists and cognitive scientists sometimes develop flowcharts to try to explain human cognitive abilities.
• By creating AI programmes based on these flowcharts we can compare the programme’s behaviour to human behaviour.
• If the programme mimics human behaviours (and accuracy), then it is likely that the flowchart is a good model for that cognitive process.
• Friedenberg differentiates between ‘computable’ and ‘uncomputable’ processes.

Gaming

• AI is used in a variety of ways in gaming, including the design and control of Non-Player Characters (NPCs).
• NPCs will be accepted more, and players will be more immersed, if these characters are designed to behave as realistically as possible.

Customer services and user interfaces

• Chatbots can be commercialised as customer service agents and user interfaces on smartphones.
• These allow for a natural interaction with the device.
• However, users may not always prefer them.
• It is possible that such chatbots could react to user mood and potentially offer coping strategies.

Psychotherapy

• Chatbots might also be useful as psychotherapists.
• One of the first chatbots was a parody of a Rogerian psychotherapist (Eliza, created by Joseph Weizenbaum).
• The Eliza effect is the susceptibility of humans to misinterpret computer output and infer meaning and understanding where it is not warranted.
• Chatbots can be used for various aspects of clinical psychological practice, such as training, assessment, clinical decision making and therapy.
• While AI therapists would have some advantages, there are also concerns.

Companionship

• One of the applications of social robotics is the provision of robots as companions, especially to the elderly and children.
• Therapeutic robots have been used in hospitals and elder-care settings in place of animal therapy.
• Virtual pets can also be included within games or smartphones apps, without a robotic embodiment. However, these have been found to be less effective than real pets.
• The physical form of the robot can affect how users interact with it.
• There are many ethical implications for robots in companionship, particularly with the elderly.
• It may happen that humans form ‘unidirectional emotional bonds’ with robots (Scheutz, 2012).
• Nevertheless, there are many potential advantages to using robots as companions.

Artificial intelligence in the future

• Philosophers, authors and researchers have considered the future of AI.
• The singularity is a time in the future where advances in AI become uncontrollable or when AI surpasses human intelligence in every way (thus becoming a superintelligence).