Judgmental selection of parameters for simple forecasting models

In an era dominated by big data and machine and deep learning solutions, judgment has still an important role to play in decision making. Behavioural operations are on the rise as judgment complements automated algorithms in many practical settings. Over the years, new and exciting uses of judgment have emerged, with some providing fresh and innovative insights on algorithmic approaches. The forecasting field, in particular, has seen judgment infiltrating in several stages of the forecasting process, such as the production of purely judgmental forecasts, judgmental revisions of formal (statistical) forecasts, and as an alternative to statistical selection between forecasting models. In this paper, we take the first steps towards exploring a neglected use of judgment in forecasting: the manual selection of the parameters for forecasting models. We focus on a simple but widely-used forecasting model, the Simple Exponential Smoothing, and, through a behavioural experiment, we investigate the performance of individuals versus algorithms in selecting optimal modelling parameters under different conditions. Our results suggest that the use of judgment on the task of parameter selection could improve forecasting accuracy. However, individuals also suffer from anchoring biases.