Mercier and Sperber's (M&S's) provocative perspective suggests that inference has adapted to a social world where argumentation is common, and that many phenomena identified as reasoning errors are not errors but adaptive when considered as inferences embedded in a social world.
We agree that inferences are often constructed when confronting a problem, and that this construction is influenced by context rather than generated by unvarying algorithms. We suggest, however, that inference construction is affected not just by social forces but also by properties of human memory retrieval and multiple goals of decision makers.
This commentary describes parallels and differences between M&S's argumentative hypothesis and a memory-based account of preference construction: query theory (QT). M&S cite two applications of QT as examples of reason-based choice with resulting choice inconsistencies; namely, the endowment effect (Johnson et al. Reference Johnson, Haubl and Keinan2007) and greater discounting of time during delay than during acceleration decisions (Weber et al. Reference Weber, Johnson, Milch, Chang, Brodscholl and Goldstein2007).
However, QT is more than another example of reason-based choice. It provides evidence and process-level specification of the implicit memory-retrieval and argument-integration processes people use to evaluate choice options and demonstrates their causal role in arriving at a decision. Just as M&S unpack intuitive inference, QT treats intuitive preferences neither as a primitive (as in economics [Becker & Stigler Reference Becker and Stigler1977]) nor as a mysterious black box (as in social cognition [Dijksterhuis et al. Reference Dijksterhuis, Bos, Nordgren and van Baaren2006a]), but instead documents the cognitive mechanisms used in constructing preferences (Weber & Johnson Reference Weber, Johnson, Lichtenstein and Slovic2006).
These are the key process specifications of QT: (1) People query past experience for evidence supporting different choice options, (2) these queries are executed sequentially, and (3) the first query produces richer representations because of output interference. This occurs because, as evidence for the first option is generated, evidence supporting other choice options is temporarily suppressed. Finally, (4) choice follows from the resulting balance of evidence. Since the order of options consideration influences the balance of evidence, it is important to know what determines which choice option gets queried first. Characteristics of the choice environment often determine what option is considered first, such as the existence of decision defaults. Like M&S, QT suggests that framing effects occur because different frames make reasons differentially available. QT finds that framing works by influencing the order in which two options are considered and thus the balance of evidence, which mediates choice (Johnson et al. Reference Johnson, Haubl and Keinan2007; Weber et al. Reference Weber, Johnson, Milch, Chang, Brodscholl and Goldstein2007). For example, different countries have different defaults for organ donation, which changes the order in which queries pro versus con donating are considered, producing different levels of organ donation (Johnson & Goldstein Reference Johnson and Goldstein2003). Similarly the order of consideration can be affected by different attribute labels that trigger positive versus negative emotions for different choice options (Hardisty et al. Reference Hardisty, Johnson and Weber2010).
Just like the inferential processes described by M&S, QT processes operate automatically, without awareness, and are effective (though sometimes biased) products of motivated cognition. The motivation for which option to consider first makes sense most of the time. Default options currently in place were typically selected for good reasons and have not caused injury or harm. Options that trigger desire have desirable features, and options that don't trigger disgust or contempt are typically superior to those that do. Giving such options an advantage by querying arguments for their selection first is a way of making the right decision faster and with greater confidence. Both inference and preference trade off between accuracy and efficiency and confidence, though these different goals do not always work in opposite directions. Whereas argumentative goals raise confidence in one's inferences or decisions and also shorten time to reach them, the initially favored options or opinions typically have good reason behind them, and the seemingly biased inference or preference reflects mostly reasonable Bayesian priors, with perhaps some built-in conservatism.
These parallels between M&S's hypothesis and QT suggest that the purpose of argumentation is not purely interpersonal, but that implicit argument recruitment, in some outcome-biasing fashion, is also an intrapsychic process that is part of implicit preference construction. Note that Franklin's comment about the human ability to “find or make a reason for everything one has a mind to do” (cited by M&S in support of social argumentation [sect. 4.1.4, para. 1]) was prompted by his internal struggle between vegetarian beliefs and the tempting smell of freshly caught fish on a sea voyage (Franklin Reference Franklin1817/2006). (He justified eating the fish by recalling the observation, while watching the fish being cleaned, that it had eaten other, smaller fish.) While this is an example of conscious inference, justification, and argumentation, M&S and QT argue that such memory retrievals and inferences occur constantly and without conscious awareness to guide our actions. Few choices offer dominating alternatives, and internal conflict between competing objectives and hence choice alternatives is the norm. Like Franklin's example, many decisions also have the potential for postdecisional regret, making it important to bolster confidence that the chosen option is in fact the best one.
Are there differences between intrapsychic and interpersonal argumentation? One difference relates to one of the most crucial components of QT's sequential argumentation; namely, the process of output interference, where the first query for evidence supporting the implicitly favored, and thus first-considered, choice option temporarily inhibits arguments for other choice options, hence reducing generation of support for it during subsequent queries. This is clearly an intrapsychic process, not an interpersonal one. It is only when I generate arguments for a given action (and not when someone else does so) that the accessibility of arguments for other actions in my mind gets inhibited. To the extent that we find similar subtle biasing of choices in line with the goals of motivated cognition in group discussion and decision settings, it must be accomplished by different processes. Smith et al. (Reference Smith, Weber, Appelt and Milch2009) suggest that interpersonal argumentation in group decisions changes goal-specific weights and not the order by which evidence is considered and thus its recall success. M&S find similar differences in implicit inference processes when inferences are made by a single individual or by several individuals in a group setting.
In summary, preference and inference tasks seem connected, among other things, by a common argumentative nature, which may suggest shared cognitive mechanisms (Weber & Johnson Reference Weber and Johnson2009).
Mercier and Sperber's (M&S's) provocative perspective suggests that inference has adapted to a social world where argumentation is common, and that many phenomena identified as reasoning errors are not errors but adaptive when considered as inferences embedded in a social world.
We agree that inferences are often constructed when confronting a problem, and that this construction is influenced by context rather than generated by unvarying algorithms. We suggest, however, that inference construction is affected not just by social forces but also by properties of human memory retrieval and multiple goals of decision makers.
This commentary describes parallels and differences between M&S's argumentative hypothesis and a memory-based account of preference construction: query theory (QT). M&S cite two applications of QT as examples of reason-based choice with resulting choice inconsistencies; namely, the endowment effect (Johnson et al. Reference Johnson, Haubl and Keinan2007) and greater discounting of time during delay than during acceleration decisions (Weber et al. Reference Weber, Johnson, Milch, Chang, Brodscholl and Goldstein2007).
However, QT is more than another example of reason-based choice. It provides evidence and process-level specification of the implicit memory-retrieval and argument-integration processes people use to evaluate choice options and demonstrates their causal role in arriving at a decision. Just as M&S unpack intuitive inference, QT treats intuitive preferences neither as a primitive (as in economics [Becker & Stigler Reference Becker and Stigler1977]) nor as a mysterious black box (as in social cognition [Dijksterhuis et al. Reference Dijksterhuis, Bos, Nordgren and van Baaren2006a]), but instead documents the cognitive mechanisms used in constructing preferences (Weber & Johnson Reference Weber, Johnson, Lichtenstein and Slovic2006).
These are the key process specifications of QT: (1) People query past experience for evidence supporting different choice options, (2) these queries are executed sequentially, and (3) the first query produces richer representations because of output interference. This occurs because, as evidence for the first option is generated, evidence supporting other choice options is temporarily suppressed. Finally, (4) choice follows from the resulting balance of evidence. Since the order of options consideration influences the balance of evidence, it is important to know what determines which choice option gets queried first. Characteristics of the choice environment often determine what option is considered first, such as the existence of decision defaults. Like M&S, QT suggests that framing effects occur because different frames make reasons differentially available. QT finds that framing works by influencing the order in which two options are considered and thus the balance of evidence, which mediates choice (Johnson et al. Reference Johnson, Haubl and Keinan2007; Weber et al. Reference Weber, Johnson, Milch, Chang, Brodscholl and Goldstein2007). For example, different countries have different defaults for organ donation, which changes the order in which queries pro versus con donating are considered, producing different levels of organ donation (Johnson & Goldstein Reference Johnson and Goldstein2003). Similarly the order of consideration can be affected by different attribute labels that trigger positive versus negative emotions for different choice options (Hardisty et al. Reference Hardisty, Johnson and Weber2010).
Just like the inferential processes described by M&S, QT processes operate automatically, without awareness, and are effective (though sometimes biased) products of motivated cognition. The motivation for which option to consider first makes sense most of the time. Default options currently in place were typically selected for good reasons and have not caused injury or harm. Options that trigger desire have desirable features, and options that don't trigger disgust or contempt are typically superior to those that do. Giving such options an advantage by querying arguments for their selection first is a way of making the right decision faster and with greater confidence. Both inference and preference trade off between accuracy and efficiency and confidence, though these different goals do not always work in opposite directions. Whereas argumentative goals raise confidence in one's inferences or decisions and also shorten time to reach them, the initially favored options or opinions typically have good reason behind them, and the seemingly biased inference or preference reflects mostly reasonable Bayesian priors, with perhaps some built-in conservatism.
These parallels between M&S's hypothesis and QT suggest that the purpose of argumentation is not purely interpersonal, but that implicit argument recruitment, in some outcome-biasing fashion, is also an intrapsychic process that is part of implicit preference construction. Note that Franklin's comment about the human ability to “find or make a reason for everything one has a mind to do” (cited by M&S in support of social argumentation [sect. 4.1.4, para. 1]) was prompted by his internal struggle between vegetarian beliefs and the tempting smell of freshly caught fish on a sea voyage (Franklin Reference Franklin1817/2006). (He justified eating the fish by recalling the observation, while watching the fish being cleaned, that it had eaten other, smaller fish.) While this is an example of conscious inference, justification, and argumentation, M&S and QT argue that such memory retrievals and inferences occur constantly and without conscious awareness to guide our actions. Few choices offer dominating alternatives, and internal conflict between competing objectives and hence choice alternatives is the norm. Like Franklin's example, many decisions also have the potential for postdecisional regret, making it important to bolster confidence that the chosen option is in fact the best one.
Are there differences between intrapsychic and interpersonal argumentation? One difference relates to one of the most crucial components of QT's sequential argumentation; namely, the process of output interference, where the first query for evidence supporting the implicitly favored, and thus first-considered, choice option temporarily inhibits arguments for other choice options, hence reducing generation of support for it during subsequent queries. This is clearly an intrapsychic process, not an interpersonal one. It is only when I generate arguments for a given action (and not when someone else does so) that the accessibility of arguments for other actions in my mind gets inhibited. To the extent that we find similar subtle biasing of choices in line with the goals of motivated cognition in group discussion and decision settings, it must be accomplished by different processes. Smith et al. (Reference Smith, Weber, Appelt and Milch2009) suggest that interpersonal argumentation in group decisions changes goal-specific weights and not the order by which evidence is considered and thus its recall success. M&S find similar differences in implicit inference processes when inferences are made by a single individual or by several individuals in a group setting.
In summary, preference and inference tasks seem connected, among other things, by a common argumentative nature, which may suggest shared cognitive mechanisms (Weber & Johnson Reference Weber and Johnson2009).