Algorithms start to get more important for companies every day. They can help companies make predictions and decisions, but since algorithms are black boxes and extremely literal, companies need to know how to manage them. The article written by Luca, Kleinberg and Mullainathan (2016) in Harvard Business Review explains more about the challenges you face when using algorithms and about how to avoid as much problems as possible. This blogpost will give you a short summary about the insights of the article.
As mentioned, algorithms are black boxes and extremely literal. Algorithms often can predict the future with great accuracy but tell you neither what will cause an event nor why. This phenomenon is called black box. Besides, algorithms are extremely literal, which means that is does exactly what it is being told. For example, when you tell an algorithm to maximize user click-throughs, websites will show low-quality click-bait articles. This leads to higher clicks, but customer satisfaction might decrease. The goal might be accomplished since user click-through is maximized, but a decrease of customer satisfaction is of course not what companies want. Therefore, it is important to know how algorithms work and how to manage them. Luca, Kleinberg and Mullainathan (2016) have set up two steps to manage algorithms better.
First of all, you should be explicit about all your goals. It is important to be clear about all the goals that you want to achieve, including your soft goals. Second, you need to minimize myopia. This means that you should not only focus on the short term (maximizing clicks or number of orders), but that you should also think about your goals on the long term. Some articles might get more clicks because they are shocking, but it could decrease customer satisfaction. It is the same for maximizing orders. Some products might sell better than others, but their return rates are also higher, which eventually does not lead to the highest possible profit. Therefore, companies should also include long-term goals.
Understanding what algorithms can’t do is as much as important as understanding what algorithms can do. Following the mentioned steps above will make it easier to interact with algorithms and use them in your business to make predictions and decisions.
References
Luca, M., Kleinberg, J., & Mullainathan, S. (2016). Algorithms need managers, too. Harvard Business Review.
