Considering artificial Intelligence to rule

It caught me by surprise when a number of people suggested that the best decision maker for government issues might be artificial intelligence. They feel that an unemotional, unbiased, purely logical decision maker would be superior. Unfortunately, true artificial intelligence as portrayed in science fiction doesn't exist.

It's confusing because many companies have added "AI" to their product definition to impress buyers. To do so, each company defines AI differently to match their product. However, these products perform relatively simple tasks that must be supervised by humans. AI is not intelligent in a way that it can solve complex issues or make independent decisions. AI is not a viable option as a governing decision maker.

Nevertheless, a university research group in Geneva, Switzerland is exploring ways that a pattern matching form of AI could duplicate citizens' votes on previous legislation. The group founder envisions a future where some form of AI could help write laws, but that appears to be Sci-Fi, not a possibility to influence our design. [1]

A think tank in the EU created an app for voters in 2019 to help them choose a candidate that matched their viewpoint. Voters answered 25 questions regarding how they would vote on past legislative decisions made by the EU parliament. Using their answers, the app matched them to candidates based on the candidates' votes. A similar app was used in the 2020 elections in the US. [2] With the tag of "AI" attached, these apps sound much more sophisticated, but they are equivalent to dating apps, comparing data in two profiles for a match.

Fundamental issues of AI for political decisions

The first risk in AI apps such as these is that the data used for AI matching is from the past. If past data is used to project a solution for the future, it may not apply. For example, if the AI program generated your vote based on your previous votes or responses, what would it do if you changed your mind? What if you would vote differently this time based on some new information or due to personal growth on your part? People and societies grow and circumstances change. If the situation is unique, past behavior would not apply and the AI projections of your choices would be erroneous.

Another risk is bias. Computer program designers have specific cultural backgrounds. The programmer's beliefs and assumptions that they accept as "truths" and "obvious" are programmed into the algorithm that controls the computer. Therefore, the type of information used, the wording of questions, as well as the design of the controlling algorithm contain the biases, beliefs, and culture of the programmer. AI software does not have its own values and beliefs, only those of the programmer.

Key takeaway: AI is not a viable governing solution. Matching previous data will never be a sound solution.