Are We Comfortable With Machines Having The Final Say? Posted on : Jun 24 - 2019

Decisions made by complex algorithms impact all areas of our lives: the ads we see, the social status updates we read, the medications we are prescribed, how much an insurance policy will cost and whether or not we get a mortgage for a new home.

Automated decisions help us cope with our fast-paced lifestyles. They are quick, they may feel relevant and they can be convenient. For example, while shopping on Amazon, you get suggestions for similar products that an algorithm has chosen based on other peoples’ purchasing habits. Most of us aren’t thinking about why we’re shown some products instead of others.

However, what about when it comes to a high-impact decision? It may not matter to you that you are shown a product that will earn the company more money, rather than a product you would actually prefer. But, say you are ill and want to secure a short-term loan to help you pay for some medical bills to get you better. You ask to secure a loan from a bank, but you are denied. You don’t understand why. You have a good credit score and a secure job with income that you can use to make the monthly payments. You go ahead and ask the bank why they denied you the loan. In many cases, the answer you get is rather vague. And in a world where we rely more and more on advanced algorithms influencing our lives and even deciding things for us, the lack of understanding is becoming of increasing interest.

The Problem With Most AI And ML

The problem with many of the newer machine learning (ML) algorithms and artificial intelligence (AI) is that they typically do not provide an explanation for the decisions they come up with. Meanwhile, we have become so accustomed to AI making decisions for us in our daily life that we’d rarely wonder about its decision-making process, assuming the algorithm is accurate and so are the decisions made.

But, this normally changes when something doesn’t go as expected. If you are rejected for a loan based on an algorithm, and there’s no human oversight to explain why, how can you be sure the decision is fair? As humans, we can accept that a doctor may be wrong 1 out of 10 times, but having a computer making the same mistake doesn’t feel quite as acceptable. And in many societies, when it comes to more complex decisions -- think of law and medicine -- historically, we have created processes to review and generate second opinions. But when an algorithm is making an unsupervised decision in less than a second, how does one know that it was right — that the information it based the decision was true and that it had the complete picture? View More