The Psychology of Failed Predictions

Psychology plays a role in every decision that we make, to choosing a mate and starting a family, to the advertisements that persuade us in purchasing the latest electronics. But, what role does psychology play when it comes to our ability to predict world events or the latest emerging technologies? Is it possible that we are limited to certain psychological barriers that hinder our ability to reasonably survey and forecast future events? In a new study by Machine Intelligence Research Institute (MIRI) describes ‘How We’re Predicting AI – or Failing To’ by Stuart Armstrong and Kaj Sotala. It shows that there is substantial evidence for suspicion when it comes to our ability to predict future events. According to the data in the study, it is possible that forecasters could struggle with subconscious psychological predispositions that hinder their ability generate successful forecast. This study was not limited to expert forecast, but also included the forecast of non-experts, and cites data that shows there is an indistinguishable difference between the predictions of expert and non-expert forecasters.

Figure 1: “Median estimate for human-level AI, graphed against date of prediction” (1).

One of the most common mistake that expert and non-expert forecasters make is the so-called Maes-Garreau Law formulated by Kevin Kelly that states, forecasters will predict that a certain event with happen within their life time. “In this case, the rise of AI that will save them from their own deaths, akin to a Rapture of the Nerds” (1), a second mistake that forecasters make is that ‘event X is within 20 years’.

Figure 2: “Difference between the predicted time to AI and the predictor’s life expectancy, graphed against the predictor’s age” (1).

In order to compile and execute more accurate forecast and predictions, it is important that we understand our psychological procedures that could obstruct our vision of the future. By doing so, it could allow us to peer into the future with unbiased eyes, and a fresh perspective of what is logically possible according to our data, and not blindly project our ego onto our scenarios and forecast.

SOURCE: (1) HTTP://INTELLIGENCE.ORG/FILES/PREDICTINGAI.PDF