The benefits of using algorithms are well documented, yet they are still not used for many decision-making situations. The reasons for this lack of use are varied. It could be self-interest. It could be algorithms anxiety. It could be a lack of confidence in the modeling process. If there is a high level of uncertainty concerning the most effective model, there may be fear of being wrong.
We know that the more structure to the problem, the less noise or error there will be with any decision. Dan Kahneman discussed the problem when he wanted to use algorithms with a group he was advising on soldier selection in the Israeli army. There was significant pushback from his client, so he came up with a compromise solution of using the set of factors involved with the decisions as a scorecard. The scoreboard of crucial success factors would be filled out with the final decision in the hands of the client given the factor information. The scorecard or checklist always led to better decision improvement than no checklist and almost as good as the algorithm.
There may be a practical middle-ground approach for any decision-making that would work well for those who do not want to turn over decision-making to an algorithm – a decision information dashboard.
Dashboards are being used more frequently in businesses across many fields. They are generally focused on providing up-to-date information on constantly changing data. This tool is perfect for investment management and can be more informative than a checklist. Bloomberg terminals have been tricked out as dashboards for years, but with new visual displays of information and flexible business analytics tools, dashboards have taken on greater usefulness and timeliness. The dashboards can incorporate the process of an algorithm in an easy-to-read format.
A dashboard can focus on specific decision-making by using graphics and scoring key variables employed in a specific decision. For example, a simple scorecard on fixed income could include the state of the overall economy, the shape of the yield curve, momentum, carry, expected inflation, and Fed policy. These scorecard factors could flash red or green based on the score. The signals can be aggregated to a fixed income buy/sell score.
The decision can still be placed in the hands of the trader. Still, the likelihood of a noisy decision is reduced since the decision-maker would have to act against the factors he believes are important. Mistakes will be made, and the dashboard may signal false positives, but the chance of misdirected actions will be reduced. The dashboard could be set signal what multiple algorithms would do. The decision-making problem could then focus on disagreement with these algorithm assessments. A feedback loop can be established for any set of decisions to match the decision-makers’ action with the dashboard score.
Dashboards can force discipline on investment decision-makers. Of course, there is a problem if the dashboard advice is not taken or proves to be inaccurate, but it is a good way to impose an algorithm-lite structure on decision-making.