AI Data Sets and Human Behaviour

AI is looking at mental health through data sets. A data set corresponds to one or more database tables, where every column of a table represents a particular variable, and each row corresponds to a given record of the data set in question. This data would then be utilized to diagnose mental disorders such as autism, bipolar, depression, and binge-drinking. In the future, data sets would provide a meta-analysis of most human behaviours, from mental health issues to potential criminal behaviour.

AI is proposing to increase the predictability of human behaviour to improve diagnoses and reduce misdiagnoses. AI is using data sets to identify consistent behaviour and neural patterns to diagnose. The question is: Are humans really consistent in their behaviour and neural patterns? Does the inconsistency of human behaviour make data sets somewhat general and nonspecific?

”Inconsistency is the only thing in which men are consistent.” –Horace Smith

Stress and Inconsistency

Stress affects consistency. We know that when stressed, even a consistent performer like an Olympic athlete may be affected. Because stress affects attention, we know that traffic accidents are more likely experienced under stressful conditions, like bad weather or running late to work. Stress is not a mental health issue but rather a normal reaction to everyday life. However, stress does appear to be part of any data set associated with human behaviour. How could it not be? What happens to the accuracy of our AI data set if stress is not one of the contributing variables measured in the meta-analysis? How accurate is the analysis?

“One cannot live without inconsistency.” –Carl Jung

Human Unpredictability

Human behaviour can be unpredictable. Humans behave in irregular, illogical, and atypical fashion from moment to moment. Calculating a logical predictable data set to any human being may be overly optimistic. One moment we are lucid and the next moment anything but. How can this moving target be labelled accurately through a column of tables? We may arrive at tendencies of behaviour but not definitive behaviour. There will always be areas of inconsistency that prevail.

Personality Inconsistency

When looking at personality, we see several examples of potential inconsistency. There is temporal inconsistency, which assumes that there is no substantial change in the construct being measured between two occasions. There is also cross-situational inconsistency, which explains inconsistent behaviours related to personality characteristics across similar situations. And, there is a personality shift that transpires between the real and the ideal selves. Which self are we dealing with? How do we separate these anomalies through database tables? How can we rely on these generalized data sets that do not include all these very specific anomalies?

A Hands-Off Human Approach

AI can help humankind by cleaning up a nuclear accident. AI can program our machines. AI can even assist humans in thinking through complex problems in a complex world. However, maybe AI needs a hands-off human approach to be most viable to humans.

We do not need to be diagnosed by machines and it seems ludicrous to think we should be. A major part of being a sentient being is our awareness of self. Why would we substitute a pseudo-model of our humanity to pass judgement on our mental health?

AI has a place in areas that we cannot match. AI has no human equivalent in the speed of processing, in the volume of processing, and even in the tedium of processing. Let AI thrive in that environment. But let’s not let AI take over the great experiment of being human. We already have that, let’s keep it.

States Not Traits

AI through data sets would be attempting to identify human traits. Traits are long-term characteristics of behaviour, actions, and feelings. On the other hand, human behaviour is also about states, which are temporary conditions that vary with the context of any situation.

The best AI can do with data sets is establish certain traits within human behaviour. We need not only tendencies of human behaviour but also the context of that behaviour (Zhang, et al., 2021). We will still have to acknowledge all the inconsistencies of human states. Data sets that look only at traits will only ever provide a partial picture of one’s mental health.