Various competing theories can describe multi-attribute decision-making, for example, the Adaptive Toolbox and Evidence Accumulation Models (Krefeld-Schwalb et al., 2017). This study takes up the idea of unifying strategies from the adaptive toolbox with the evidence accumulation approach (Lee & Cummins, 2004). Published data is reanalysed (Bobadilla-Suarez, 2017; Bergert & Nosofsky, 2007) by applying a linear decomposition of the drift parameter of the Drift-Diffusion Model (Ratcliff, 1978) in a weighted strategy-specific component (validity rank of the first discriminating attribute for Take-The-Best; difference of positive attribute values between options for Tallying).
We replicated the use of a strategy by the likelihood values of models with different strategy injections. Further, the graphical goodness of fit was satisfactory because the fitted and observed behavioural data overlapped. Additionally, we showed that the model maps strategy-specific difficulties: For the Take-The-Best strategy, the model mimics slower and less consistent responses when the search depth for the first discriminating attribute shifts to the lower end of the attribute matrix. Identical results are found for Tallying by varying the difference of positive attribute values as strategy injection.
An advantage of this method is that we can simulate reaction times depending on the environment and the strategy usage. So far, it was only possible to contrast reaction times between different strategy-specific difficulties (Jekel, Fiedler and Glöckner, 2011).
We could mimic strategy-specific observed behaviour by a very rudimentary strategy-specific injection into the drift parameter. Further research will investigate investigate whether injections on, e.g. the boundary/non-decision time parameter will improve the model.