Choice probabilities and related outputs of discrete choice models form a critical input to many travel demand forecasting and transport project evaluation studies. The decision rule underlying a discrete choice model describes how individuals make their decisions and thereby co-determines the choice probabilities. Uncertainty from the side of the analyst regarding the underlying decision rule(s) may therefore translate into alternative predictions regarding the behavioural response to changing travel conditions. In this paper, we contrast the well-known Random Utility Maximization framework, on which most travel demand forecasts are based, with its Random Regret Minimization counterpart. Based on a review of the existing empirical comparisons between the two frameworks we discuss the connections and dissimilarities between both model types and the associated implications for travel demand forecasting. The empirical comparisons reveal that both models perform about equally well in terms of model fit and external validation, which makes it hard to identify one model as a superior specification for forecasting. Despite these small differences in overall model fit, choice probabilities and elasticities can differ substantially (and predictably) in specific choice-contexts. One such example is the compromise effect where the Random Regret Minimization framework predicts a market share bonus for ‘in-between’ alternatives. The paper discusses model averaging techniques to generate predictions when a clear winning model cannot be identified. Finally, the paper puts these considerations in the context of a regret-based Dutch National Model, which is currently under construction.