In spite of the strong growth of systematic trading, the problem of “algorithm aversion” is real and must be addressed. In many cases and in many disciplines, models do better than humans, but humans feel anxious with models and do not want to place decision-making in the hands of a machine. Even if a model does better at forecasting, individuals may still prefer the discretionary or non-alto approach if there is the perception that the model is imperfect. It seems as though individuals will like to have the optionality to choose an approach (model or discretion) as opposed to being locked into one choice. See our posts: “Algorithm aversion” and managed futures and Algorithm aversion or just a desire for low cost optionality

The same researchers who identified algorithm aversion have published new work which suggests simple ways to avoid the problem. (See Overcoming Algorithm Aversion: People Will Use Imperfect Algorithms If They Can (Even Slightly) Modify Them by Dietvorst, Simmons, and Massey) The authors find that if some modest amount of control is given to the decision-maker, he will choose the algo over his own decision-making. Allow the individual to adjust the model output by 10% and they are happy. Allow the individual to reject 10% of the output from the model and they are happy. Aversion falls when the human has some control over the model. You could say that the way to combat one form of behavioral bias – algorithm aversion can be through another bias, the illusion of control. Fight the bias with a bias. Or, allow the decision-maker to have some optionality to override and aversion goes away.

I have seen this happen in the systematic model space. Investors seem to be always fearful that the manager who always follows the model will at some point drive it into a brick wall. If the decision-maker has, under extreme circumstances, the chance to turn-off the model, anxiety falls. The optionality does not have to be used. It just has to be available. This can be tested through real life survey. If an investor is shown two systematic managers, one with no discretion and the other with some modest discretion, the question can be asked as to whether one will be preferred to the other. If there is a preference, an additional question can be asked as to the amount of excess return necessary to switch. Investors do face this type of problem.

Algorithm anxiety affects investment decisions, but research suggests that it can be overcome so that the best choices are made in spite of biases. This is especially important given the backlash against systematic managers that is starting to creep into news stories.