Automated Curiosity - Breaking the Bubble Automated Curiosity - Breaking the Bubble

By Antti Hahto


In our previous articles in the A.I. series, we established some background in Pattern Recognition (Go: Post-Mortem Humanity) (Pattern Recognition: Prologue) and introduced a connection to User Experience Design.

Now, a question: how do you make informed decisions? By obtaining the information first, of course. I don’t believe in intuition or any other magical property of the human mind. Expressions like “it just feels right” take too much credit away from you. Don’t be so modest.

Not so surprisingly, one’s so-called intuition gets better with practice, no matter what the subject is. Practice is simply the process of trying things over and over, while making adjustments to get better results. These adjustments come from feedback reinforcement, and from continuously integrating new information from various sources. Getting good at something (or some lucky ones already having it primed in their genes) feels almost automatic. That’s intuition.

Algorithmic bubbles are currently a hot topic – and one that’s not going away any time soon – so be prepared. Just by reading this article, your interest in the subject will be added to your digital profile by some of the most profitable companies on the planet – at least if your web browser has cookies turned on.

The bubbles form as we gradually offload our decision-making and information gathering to third parties. While looking at flights to Munich to visit Intopalo’s cool new office, or browsing the movie catalog on Netflix, it’s not you who makes the decision. Rather, algorithms determine the price of your potential flights and the 1-star dystopian horror movie you might want to watch. If you think you’re the one in control of the decision, consider this: a husband may make the final decision, but it’s the wife who provides the options. Of course, you might make some decisions outside the given options, but those might not be the informed ones, and they may require extra effort. And let’s face it, what wouldn’t we do for some extra convenience?

As for the bubble part: since the information available to you is limited according to certain criteria, you mostly make decisions and gather knowledge within those limitations. Be it political views or movies, this fact only works to further reinforce your soapy, comfortable walls. Moreover, you might not even like dystopian soap operas – maybe you just happened to watch a commercial and do a web search.

In a business setting, offloading decision-making from people to machines can be a good thing. This is mainly because humans are horrible at making decisions. According to research, the outcome of a parole hearing is affected more by whether the judge ate lunch prior to the trial than whether the person was actually guilty. The study in question found that the odds for a prisoner to be successfully paroled started off at around 65%, plummeting to almost zero over a few hours. After the judges returned from their lunch breaks, the probability climbed back to 65%.

Most companies are still run by Excel, which is analogous to predicting stock prices by technical analysis with known formulas and attempting to outcompete others doing the same. Despite the abundance of data available, companies are still using the same-old approach of linear regression lines plotted over sales forecast figures. Don’t do it.

Letting an A.I. wife make one’s personal decisions might eventually make life miserable. In the business world, however, competitive advantage comes from processed knowledge. Although the decision-making abilities of humans can be questionable, three 2017 Nobel laureates would agree that the brain’s risk-reward pathways are a pretty good model to imitate: in its simplest, future behavior is dictated by a constant feedback loop based on whether anticipated rewards were delivered as expected or not. This link between reward prediction error and actual reward has been, maybe surprisingly, one of the biggest steps in linking A.I. to economics and engineering.

In Miller’s Information Processing Theory, humans can be seen as computers. This is somewhat ironic, considering computers to become more human. In this article, we didn’t get into the automated decision part, yet. Frankly, we didn’t even get to the information gathering part, being very much related to the role of attention and furthermore, constructing an actual Expert System for running your business. So you don’t have to.

Set goals, learn from history. Be amazed.