For decades, context effects have been a primary focus of investigations into multialternative, multiattribute decision making. Originally studied to demonstrate violations of classic decision theories, in recent years context effects have helped researchers investigate the cognitive and neural mechanisms underlying multialternative, multiattribute choice. As discussed in this review, there are many different theoretical accounts of context effects. Although attention is common to most dynamic models of context effects, divisive normalization offers an alternative explanation. It is possible that there are multiple pathways to context effects. This might explain why context effects often appear to be ubiquitous, occurring in different domains (Trueblood et al., 2013) and species (Latty & Trueblood, 2020).
An important outstanding problem is understanding reversals of context effects. Recent studies have documented such reversals (see Spektor et al., 2021, for a review), but it is unknown when and why they occur. On the one hand, this has caused some researchers to question the importance of context effects (Frederick et al., 2014). On the other hand, reversals provide researchers an opportunity to refine their understanding of how fundamental cognitive processes, such as attention, play a role in multialternative, multiattribute choice. It will require both empirical and computational-modeling work to elucidate these processes.
On the empirical side, new experiments are needed to explore the factors that lead to reversals of context effects. I hypothesize that attention might be critical to understanding reversals, given the research showing that presentation layout influences the effects. However, it remains unclear exactly how presentation layout affects attention. Eye-tracking studies that include manipulations of layout could help shed light on this question.
On the theoretical side, research is needed to understand the cognitive mechanisms responsible for context-effect reversals. Many of the existing models discussed in this review cannot fully account for recent empirical demonstrations of context-effect reversals or how choices change over time (Cataldo & Cohen, 2021). Thus, new modeling approaches are needed. The switchboard modeling approach of Turner et al. (2018) could prove very useful in exploring models that can produce both standard and reversed context effects and in identifying the relevant underlying cognitive mechanisms. This approach could help distinguish which of the various implementations of attention listed in Table 2 accounts best for behavior. Another important goal for future research will be to evaluate models using information about how choices change over time. To date, there have been only a few studies evaluating models using both choices and response times (e.g., Cataldo & Cohen, 2021; Evans et al., 2019). In sum, coupling the new modeling approaches and empirical studies aimed at manipulating the presence, reversal, and even absence of context effects could reveal new insights into the dynamics of multialternative, multiattribute decision making and how preferences evolve over time.
