Can AI Make Us More Rational?
How AI-aided choices can make us more rational, for better or for worse
July 21, 2019, Bruno Jacobsen
People are irrational. Evolution has shaped us that way. Almost everything we do involves some sort of decision. Most of the time, we are unaware of them. Our system 1 performs automatic, fast, and often unconscious decisions. It's also prone to bias and judgement errors. Even when we engage our system 2, and think hard about a problem or situation, it is often difficult to shake off system 1. We don't usually brake down problems into all its constituents, down to each and every assumption. Fortunately, artificial intelligence can be our friend - let's look at different ways in which it will help us with decision making in the future.
Can AI Move Us Past Irrational Behaviour?
Letting algorithms make decisions for us might not appeal to everyone. An algorithm is just a set of rules that a person or a computer follows when trying to solve a problem. What does it know about anything? When it comes to real-life decisions, aren't we better off doing some real thinking? Are we taking in all our experiences, our knowledge, our intuitions, and making a decision then?
Of course, the obvious answer to that is "no." And most of us know that. You need only to look at Google DeepMind's AlphaGo, beating 18-time Go world champion Lee Sedol. While Go may not resemble every-day life decisions perfectly, it nonetheless illustrates something obvious: algorithms can make better decisions.
And we know they can also help us in daily life. Ray Dalio, American hedge fund manager and billionaire, in his book Principles suggests writing down the criteria we use to make decisions and go back to them after the decision has turned out to be good or bad. Once an algorithm has proven to work under certain circumstances, then we should always use it. These are simple "if X, then Y," with X being the criterium (or criteria) and Y being the action taken.
Taking this into consideration and understanding the fallibleness of human judgement, we can begin to understand how AI may help us. And it can help us a lot when it comes to our own biases.
For instance, in-group bias is a well-known bias in which members of a group tend to make more favourable judgements about members of the same group than about members of other groups, or "out-groups." As much as we like to believe many of us do not suffer from this bias or at least do not act on it, it frequently happens unconsciously. It is also often brought up during conversations about prejudice, whether we're talking about hiring people of different ethnicities or different gender.
AI can analyse data with more objectivity; while leaving attributes such as race and gender out of the picture.
Attentional bias is another bias that AI can help us overcome. We are often led by the thoughts that spontaneously occur to us. We may think we can control them, but as far as we know, they mostly pop into existence. That means we can overlook essential things or fixate too much on a stream of thought. By using AI designed to consider problems from multiple perspectives, this too can be overcome.
Another pervasive bias is confirmation bias. When we make a decision or have a belief, we tend to notice information that supports our beliefs much more than the ones that go against them. This confirmation bias can lead to overconfidence, and ultimately to poor decision-making. By programming an AI that understands underlying assumptions and tests them against relevant data, we can improve our decision-making. For instance, let's assume you believe most schooling and education in the future will happen online at home, not in classrooms. If you're an entrepreneur or working in an organisation that deals with the education industry, you might make decisions based on this belief. After all, everyone you know is talking about online courses, online certificates, YouTube lectures. It seems obvious. What if you were to feed this belief to an AI and it came back with overwhelming evidence that the trend is actually towards face-to-face learning and in-person classes at learning institutions?
Or we can talk about negotiations. Pick up any book on the subject, and it will teach you some fundamental biases you can use in your favour. For instance, the anchoring bias. It goes something like this: by being the first to throw down a number at negotiation, you can anchor the other person to that number. The counter-offer you'll get will likely be closer to the number you said than it would have otherwise been. Of course, well-seasoned negotiators might not fall for this, or may even use it against you. But negotiators are nevertheless extremely prone to biases, emotional outbursts, and other factors that impair good decision-making.
It's not too far-fetched to imagine that an AI, when given a set of parameters, the weight of different options and data, will make more "rational" decisions without ever getting emotional. Will it come out with win-win outcomes? Maybe. Will it be perfect? Maybe not. But the potential is there.
But we should exercise caution
While AI has the potential to make us more rational, we must also ask: can AI be rational? If we consider a set of algorithms that always makes the "right" algorithmic decision based on the set of information it has, then it could. It's hard to imagine; after all, AI scraping the internet for things that confirm its "decisions."
Unless, of course, we program it to do that. And that's where the key point lies. According to this MIT Technology Review article, AI bias isn't just a concept - it exists. The article shows that there are three stages where it could occur: when you are framing the problem, collecting the data, and preparing the data. Humans still do all of those. It doesn't matter how well you train the AI or how good your machine learning algorithms are (well, it matters a little). For example, Amazon's algorithm used previous data to train its AI. Because in that data, there were more instances of men being chosen over women than the other way around, it learned to do the same.
The point here is not to say AI cannot be more rational or help us make more rational decisions. But to illustrate the problems with AI decision-making that we haven't solved yet.
Nevertheless, for many decisions that involve large quantities of data and that are not as politically and socially charged, AI can still help us make better and faster decisions. In many cases, it has already automated our decision making (dynamic pricing, used by many companies today, is one example). In the future, we can expect AI to continue creeping into the realm of decision-making. Someday, it could even begin taking over higher-level and more creative decisions. We're not close to that, but it seems like we're not that far either.
Using AI to Enable Future Thinking
The job of a futurist involves examining and testing different possible outcomes–potential threats, emerging ideas, exciting opportunities on a daily basis.
And with the overwhelming amount of new information every day, it is a challenge to navigate through the noise and find the right signals.
That's where AI and machine learning come in.
At Futures Platform, we use AI & machine learning to find the latest updates on specific trends and phenomena. The AI-Powered Search also uses an algorithm to fetch relevant themes and topics based on the search keywords.
Constant News Updates Thanks To An AI bot
Most people who work as professional futurists know that you never stop gathering information. New data come in regularly, and you never know when a provocative, potentially disruptive development might appear.
The Futures Platform AI bot crawls thousands of articles every day and fetches the most relevant news. Reliable news sources particular to the issue are of course, essential so that our bot only uses articles from highly reputable journals or Web sites. You can find more information about Futures Platform Source Reliability Guideline here.
Search For A Particular Subject With The AI-Power Search
When you have a question to which you don't have an answer, what do you do?
Go to Google. Of course.
In a way, Google is the largest encyclopedia in the world with billions of webpages. However, because of its enormous amount of data for a given topic, it's not a simple task to find high quality, highly relevant content.
So we ask ourselves the question "Can we build our own Google for future phenomena - with high quality, tailored-made content?"
And that's why the Futures Platform AI-Power Search was born.
With the AI-Power Search, you can search for hundreds of trends and phenomena. All of our content is created and reviewed by a team of leading future experts, assuring the content quality and reliability of the highest standard.
Want a sneak peek at the Futures Platform tool? Free Registration HERE.