Resisting knowledge,realizing values,and reasoning in complex contexts: Ecological reflections

Feminists and psychologists: The value of clarity

Although experiments in thinking and reasoning are often designed as if they were Cartesian isolation chambers, experiments are actually conversational dialogues between experimenters and participants. A classic example is Tversky and Kahneman’s (1983) Linda problem. Linda is 31 years old, single, outspoken, and very bright. She majored in philosophy. As a student, she was deeply concerned with issues of discrimination and social justice, and also participated in antinuclear demonstrations. Participants were then asked which of the following two alternatives was more probable? (a) Linda is a bank teller, or (b) Linda is a bank teller and is active in the feminist movement (p. 299). The majority of people who were asked this question chose the second option, a logical error according to Tversky and Kahneman. The probability that Linda is a bank teller and a feminist cannot be any greater than the probability that she is a bank teller.

Hertwig and Gigerenzer (1999) offered an ecological and pragmatic critique of this claim. The key terms in the Linda question that create difficulties are the terms and and probable. Tversky and Kahneman assumed, without showing that it is warranted, that the word and is understood in its logical operator sense. People are unlikely to understand it in this restricted sense, but in the more typical sense, of “and here is further relevant information on which to base your decision.” Regarding probable, most people conjure up meanings such as possible, typical, reasonable, or plausible. If one rewrites the Linda problem to be more explicit about what is meant by these terms (e.g., using natural frequencies), the percentage of people answering the question incorrectly drops from about 80% to l0% or less (Hertwig & Gigerenzer, 1999).

From a values-realizing perspective, what Hertwig and Gigerenzer (1999) have shown is that the Linda problem is not clear, but ambiguous. It appears that pragmatic insensitivity by psychologists may have been misread as the intellectual insensitivity of participants. This claim is reinforced by Stanovich and West’s (2008) finding that performance on the Linda problem and general intellectual ability are relatively independent, and by developmental researchers’ finding that 8- to 10-year-old children nearly always answer set inclusion (i.e., conjunction) problems correctly (Reyna, 1991). The Linda problem was deliberately created to be complex (i.e., irrelevant information that invited an inference of relevance), leading participants to attempt to be comprehensive (e.g., determine implications for action), by assuming the issue was one of social norms (i.e., understanding other people) rather than logical norms. Given the lack of clarity and coherence just described, the complexity offered does not enrich participants’ understanding but frustrates it. Complexity without clarity and coherence undermines comprehension.

Finally, would people improve their performance on this problem if it were discussed in a group context, along the lines suggested by Mercier and Sperber (2017)? This has not been addressed to my knowledge, but it is not clear that conversation would help much, since the ambiguity and incoherence of the question would remain.

Squeaky mice: The value of coherence

Although the Linda problem was just used to highlight the value of clarity in reasoning, it revealed that coherence, comprehensiveness, and complexity are involved as well. To illustrate the importance of coherence (i.e., experimental procedures that cohere with the lived concerns of those whose thinking is being evaluated), as well as the need for an argumentative context of care, consider the famous four-card problem of Wason (1968). Cards presented might show A, 6, D, and 7 on the visible side. Participants are instructed as follows: If a card has a vowel on one side, it must have an even number on the other side. Which cards must be turned over to determine if this true or false? As it turns out, only a very small percentage of people identify the relevant cards (A, 7). The conclusion often drawn from this finding is that people are lousy logical thinkers; however, subsequent research has painted a quite different picture (Manktelow, 2012). If the task is made to be socially and pragmatically meaningful (e.g., it has implications for action and coheres with prior experiences), most people answer correctly. Numerous experiments have embedded the four-card problem into social, moral, and prudential contexts that have yielded “logical” answers. Examples include: If a person is drinking beer, then the person must be over 19 years of age. The four cards are: Drinking beer; Drinking lemonade; 22 years old; 16 years old (e.g., Griggs & Cox, 1982). If you [a nurse] clear up spilt blood, then you must wear rubber gloves, and [parent speaking to child], If you tidy up your room, then you may go out to play” (Manktelow & Over, 1991). Cummins (1996) has shown even young children can do the task successfully: It is not safe outside for the squeaky mice, so all squeaky mice must stay in the house. Yes, all squeaky mice have to stay in the house.

All of these studies and others like them suggest that it is only in the real, lived contexts of responsibilities and obligations, of authority and permission, of causes and consequences, that it makes much sense for humans to talk about reasoning. When safety is at stake (e.g., spilt blood, squeaky mice), or truth, freedom, and other values are on the line (e.g., the tidy room yields the promised freedom), then adults and children seem quite capable of testing the validity of relationships. From an ecological, values-realizing perspective, people involved in a Wason task do not engage in reasoning as an exercise in cognitive gymnastics. Rather, they explore possible ways of testing the integrity of people’s actions and relationships within events. The hypothesis offered by values-realizing theory is that values provide motivation to care, as well as criteria by which to assess how caring and careful reasoning is (Hodges, 2009).

Reasoning is improved not only by posing problems that cohere with the cares and concerns of participants; it is also improved by creating dialogical contexts in which problems are discussed. Performance on the Wason task, even the difficult vowels and numbers version, is remarkably good when a group is required to discuss it and develop a common answer (e.g., Mercier et al., 2015; Moshman & Geil, 1998). An ecological, values-realizing approach takes reasoning to be fundamentally social and dialogical: individuals reason together argumentatively in order to understand what is going on and what should be done. When presented with clear problems that cohere with participants’ experiences and cares, reasoning, particularly if it is carried out argumentatively in a group setting, is generally quite effective. Children and adults seem to be able to reason together critically and carefully much of the time (Mercier, 2011b), and these skills are not limited to particular cultures (Boku et al., 2018; Mercier & Sperber, 2017).

Framing effects: The value of comprehensiveness

One of the most important lessons emerging from the many studies done on Wason selection tasks has been that the quality of reasoning is altered by the way in which the task is framed (Manktelow, 2012). Implicitly or explicitly, selection tasks make reference to values, such as safety, truth, and freedom, which provide the motive for caring about the rule that is being tested. Framing effects, per se, have been widely studied in psychology and economics, but usually for the purpose of demonstrating biases of one sort or another (Kahneman & Tversky, 1984). Experimental designs present ostensibly identical information to participants but, framed differently, often lead participants to make different choices or draw different conclusions (e.g., people prefer 75% lean meat to 25% fat meat). However, what appears to be a logical error may not be. If I am visiting my doctor to discuss a possible operation on my knee, does it matter if the doctor says that the operation has an 80% success rate versus framing it in terms of a 20% failure rate? Sher and McKenzie (2008) noted that “framing manipulations [are generally viewed] as implanted bits of knowledge rather than as informative utterances issued in a communicative situation” (p. 91). However, in a conversational context, if the doctor says my chances of being worse off are 20%, I do not take that to be equivalent to her saying the success rate is 80%. In a human interaction in which one trusts doctors to provide advice about one’s health, and to be knowledgeable about better and worse options, it makes sense to pay attention to their subtle, pragmatic hints about better and worse courses of action. Furthermore, Sher and McKenzie (2008) observed: [The speaker said A] is not logically equivalent to [A], and it most definitely is not equivalent to, [The speaker said B], even if A and B are reciprocals of one another. Framing experiments often do not take account of the larger, more comprehensive, communicative context in which judgments are made; the framing is too small. Evidence suggests such considerations matter: if there are good reasons to doubt the trustworthiness and care of the person doing the framing, effects are reduced or eliminated (McKenzie, Liersch, & Sher, 2011).

Comprehensiveness is also shortchanged when information is isolated from its historical context. McKenzie (2004) put a full and an empty glass on a table, and then asked participants to pour half of the full glass into the empty glass, then to place the half full glass at the edge of the table. After the pouring, either glass was half full: Which did participants move to the edge of the table? Most chose the previously empty glass, and if they were asked to move the half empty glass, they chose the previously full one. Pragmatics takes dynamics into account—glasses of water are events, not static objects. They have histories and futures that are relevant to action. Questions can be better answered, and actions more effectively ordered, when understood in a more comprehensive, dynamic, and developmental context.

Despite pointing to limitations in the way researchers sometimes frame their experiments on framing, such effects appear across a wide array of tasks and domains, and sometimes can have considerable practical significance. The question that a values-realizing, ecological approach poses is, what are good ways to frame framing effects, both in research and in terms of practice (Keys & Schwartz, 2007)? Thinking about this question in terms of comprehensiveness suggests a hypothesis: often it might be wise to use multiple frames to evaluate important choices, policies, or procedures. As an example, it has been argued that science benefits from different formulations (framings) of the same physical law: “Psychologically … they are completely unequivalent when you are trying to guess new laws” (Feynman, 1967, p. 53).

If one considers the issue of framing from the perspective of argumentative discourse, different framings of some tasks are more likely to emerge in a community than in an individual. Research has shown that if there is group discussion of a problem in which one frame (e.g., positive) is adopted in a first session and a different frame (e.g., negative) in a second session, this sometimes moderates the decisions of the group so that they are more balanced (Paese, Bieser, & Tubbs, 1993). On the other hand, if the same frame is used for both sessions, decisions can become more polarized. Although the argumentative reasoning thesis (Mercier & Sperber, 2017) claims that diversity of viewpoints (e.g., frames) within a group is important, Kuhn, Floyd, Yaksick, Halpern, and Ricks (2018) found that multiple discussions of a complex issue (e.g., capital punishment) were beneficial, even when people shared the same general position. They found that people altered their arguments in response to what they heard from others, so that most participants adopted stronger arguments and dropped weaker ones over time.

Inconsistency in moral thinking: The value of complexity

As the comprehensiveness of frames of reference are extended, the complexity of reasoning, understanding, action, and character come increasingly into view. “Consistency is often regarded as an end in itself” (Sher & McKenzie, 2008, p. 91); however, such concerns need to be placed in a larger epistemic context, one that requires what Aristotle described as phronesis, not just episteme. Virtue and wisdom, not just technical skill and intelligence, are required to think rightly and act adaptively (Manktelow, 2012, p. 249). Despite this call for greater complexity by some, others have advised policy makers and ordinary citizens to use cost-benefit analysis (CBA) in making decisions. Too often, the argument goes, people let emotions (e.g., a disgust reaction to incest) or a commitment to some moral rule (e.g., do no harm) distort their judgments, leading to outcomes that are described as “suboptimal” (e.g., Bazerman & Greene, 2010).

The following example based on Paharia, Kassam, Greene, and Bazerman (2009) illustrates. A pharmaceutical company provides a valuable drug for cancer patients, but the company’s profit from selling the drug is very small. Increasing the price of the drug is very unlikely to reduce demand for the drug, although it will create financial hardship for some patients. The company decides to triple the price of the drug. Ethical or unethical? Not surprisingly, most people say unethical. The problem from a CBA perspective arises with an alternative scenario in which the pharmaceutical company sells its rights to the drug to a smaller company for 10 million euros, and the smaller company, in an attempt to recoup its investment costs, raises the price to five times the original price of the drug. People presented with this scenario give it unethical ratings too, but slightly less so than for the first scenario. There is no logical problem here, but Bazerman and Greene (2010) argued that there is a moral problem. Based purely on negative consequences for the patients using the drug, the second scenario is worse than the first, so it should be judged more harshly. The basis for their argument is that, if people are presented with both scenarios at once rather than one at a time, they judge the second scenario as more unethical. Bazerman and Greene’s moral bottom line is that people should be consistent across integrated and independent judgments, and integrated judgments should set the standard for the independent judgments. Giving priority to integrated judgments makes sense but ignores the fact that events unfold over time, and reasoning is an ongoing process. Furthermore, consistency as the criterion for what is moral is dubious. Inconsistent decisions are often both coherent and clear, if they are viewed in a larger, richer judgmental context (as was demonstrated in the Asch, 1956, study described earlier). From the perspective of values-realizing theory, isolating a single value (i.e., health) is problematic. Judging how that value is realized based on consequences alone (or intentions, or any other isolated dimension) is even more problematic, especially when only one particular consequence is included in the theoretical analysis.

The problem with CBA is that the comparisons offered by researchers to their participants (e.g., the two scenarios about the cost of a vital drug being increased) are carefully selected to screen out an enormous variety of other considerations the researchers deem “irrelevant” to a proper moral judgment, and to make salient the particular dimension they think is relevant. In short, the problems are framed in a way that biases decision-makers (e.g., which scenario increases prices more), providing a tunnel-vision perspective on the moral landscape. These “closed-world assumptions” are unrealistic and it is clear that participants in decision experiments regularly reject them (Bennis, Medin, & Bartels, 2010). Instead, decision makers try to open up their considerations of the problems that they are offered to a larger range of social and moral dimensions. CBA researchers have not offered a rationale for closing down the world in this way, and Schwartz (2010) has argued that the difficulties with thinking in terms of CBA are “not (just) in us, but in the world” (p. 205). Overall then, CBA attempts to clarify moral judgments, but it does so by shortchanging complexity, comprehensiveness, and coherence. By refusing to consider the whole array of moral resources offered by biology, history, and community to help guide choices, and by oversimplifying the openness and complexity of the world, CBA offers a clarity that turns out to be largely illusory.

Another excellent example of the importance of complexity for locating reasoning in a richer ecological context is the sunk cost effect, which is the tendency to continue to incur costs in a venture even though the prospects are not good (e.g., Although I don’t feel like going to a concert tonight, I go anyway because I had earlier purchased a ticket for $95). Economists routinely treat this as an error; however, Mercier and Sperber (2017) have suggested that following through on commitments one has made, even when it is later learned that it will prove more costly to complete those commitments than previously imagined, sends strong signals to others that the person is someone who is dependable and reliable. Although taking account of sunk costs may look incorrect when it is understood as an isolated choice, it can say something powerful about social solidarity, truth, and trust in the larger context of a community. Others have proposed additional legitimate reasons for taking sunk costs into account (e.g., Keys & Schwartz, 2007; Polonioli, 2016).

A particularly intriguing example of resistance that puts knowledge in a larger, more complicated moral context is information avoidance (Gigerenzer & Garcia-Retamero, 2017; Sweeny, Melnyk, Miller, & Shepperd, 2010). Hertwig and Engel (2016) offered a range of reasons people sometimes choose not to know facts relevant to their lives (e.g., gender of their unborn child). One of the most interesting is that people avoid information in order to increase their sense of agency—providing themselves with greater opportunities to act with confidence and hope. Hertwig and Engel (2016) have suggested that deliberately ignoring relevant information may be crucial to taking on an ambitious project: “It is possible that no textbook would ever be written, no house built, and no opera composed if people based their decision on the progress and success of similar endeavors” (p. 362). Hansson (2010) has argued that the very act of assigning probabilities to various possibilities has moral dimensions that cannot be ignored. What if, he asked, people treated national averages about the probability of divorce as a prediction of the likelihood of their own marriage failing? The frequency of divorce might well increase.

These and other examples illustrate that good reasoning often requires more than straightforward problem-solving or decision-making. Questions about what information to acquire, whether past decisions should affect current decisions, and if, when, and how to make cost-benefit analyses are among the many issues that must be decided or assumed (Keys & Schwartz, 2007). Even when someone has decided to apply cost-benefit reasoning, crucial and controversial issues loom: What is to count as a cost or a benefit? Whose costs and benefits are included or excluded (MacIntyre, 2016)? A more ecological account of values, such as values-realizing theory, can contribute to engaging in such evaluations, but such a theory claims that there are no rules or natural laws that make it easy to do what is wise (Hodges, 2009; Hodges & Baron, 1992).

Theoretical prospects

The first part of this article presented evidence and arguments for the interdependence of truth, social solidarity, and trust, and the importance of resistance and argumentation to the nature of reasoning. The second part made the case that reasoning depends on the values of clarity, coherence, comprehensiveness, and complexity mutually constraining each other so that actions and decisions are marked by integrity and character. One of the key claims of ecological values-realizing theory is that actions and decisions always involve multiple values. Thus, at any given moment, choices about resisting others or agreeing with them are not directed by predetermined goals or by maintaining consistency, but by working out the tensions between mutually constraining values.

Taking into account the research and theory described earlier, what might be offered as strands of what could become a more substantial theory of ecological reasoning? Here are some working hypotheses: (a) reasoning is primarily a social, dialogical, perception-action skill, rather than an individual cognitive skill; (b) a primary means of reasoning is to engage in cooperative, argumentative conversation; (c) in most ecological contexts, reasoning cannot be judged by accuracy, consistency, or any other isolated value or virtue; (d) reasoning cannot be done properly outside of a larger context of care; (e) reasoning is judged by the same criteria that judge the worth of our conversations and the character of our communities (i.e., clarity, coherence, comprehensiveness, and complexity); (f) reasoning requires community and common ground, but also resistance, all of which require a commitment to truthfulness, trust, and social solidarity; and (g) reasoning needs to be judged less with respect to isolated decisions, and more with respect to the direction of judgment and action over time.

In their analysis of reasoning, Keys and Schwartz (2007) concluded that what experimental researchers most need in order to make progress is a substantive theory of purpose within which activities of reasoning can be understood. Although no such theory is on offer, values-realizing theory, along with other ecological and social approaches to reasoning illustrated in this article may be small steps in the direction of such a theory.

Methodological markers

Although weaknesses in a number of important lines of research were identified earlier, the focus of those concerns was less about methodology and more about the theoretical assumptions and interpretations tied to those methods. The point of an ecological approach is not to “get out of the lab and into the real world” but to have lab or field research that is sensitive to the values that constrain choices and actions, rather than assuming a set of rules that reasoning must conform to in order to be considered appropriate.

Based on values-realizing theory, several methodological guidelines emerge. First, reasoning is rarely if ever governed by a single goal or guided by a single value. Driving is not just about being safe, or accurate, or fast, or comfortable: it is about all of these values and more (Hodges, 2007; Hodges & Rączaszek-Leonardi, 2019). The same is true of reasoning. Thus, researchers would do well to consider multiple goals and multiple values. Second, there are no simple rules that define good reasoning. Incorporating sunk costs into one’s thinking may often help to build character and reputation, as noted earlier, but if one always doubles-down on prior investments, this may well show a lack of insight and wisdom. “Inconsistency” over time may be precisely the best way forward. Enlarging the criteria for good reasoning beyond fixed rules will allow researchers to appreciate better the diversity of situations participants face. Third, research tends to study isolated decisions or judgments, such that the temporal aspect of reasoning is often lost to view (Keys & Schwartz, 2007). As noted earlier, errors in answering individual questions in the Asch dilemma looked quite different when viewed as parts of a larger whole (Hodges & Geyer, 2006). Studies that include a temporal or developmental dimension, or that place decision-making in a social, argumentative context are likely to provide a more accurate understanding of both the strengths and weaknesses of reasoning. Fourth, although any one study is limited in the variables that can be included, across a series of studies, researchers can work to address a broader array of contexts, goals, problem types, and environments in studying how people reason. It is easy for researchers to take the first version of a paradigm (e.g., Wason card task) as the fundamental phenomenon and to see later findings as exceptions or qualifications, rather than determining whether the larger picture challenges the original assumptions in crucial ways.

Much of the research discussed earlier has illustrated some of these suggestions about how to take a more ecological approach to research on reasoning (e.g., Hertwig & Engel, 2016; Moshman & Geil, 1998). Another, quite different, example is found in Erev, Ert, Plonsky, Cohen, and Cohen’s (2017) recent study of decision-making. It offers an intriguing example of a research program that embodied a social, quasi-argumentative methodology (i.e., a competition among researchers), focused on a broad set of problems and phenomena, which were taken from the reasoning literature that has used formal models (e.g., maximum expected returns on monetary gambles). They created a competition in which researchers worked to develop the most accurate model that could predict results across 14 different types of problems (e.g., economic choice games) that had previously revealed a wide variety of biases (i.e., departures from maximizing available payoffs). In short, they developed a comprehensive array of experimental problems that had been designed to provide clear tests of formal reasoning principles (maximize monetary outcomes) while accounting for complexities (i.e., biases) in a coherent way (i.e., formal predictive model). Their approach challenges the tendency of researchers to focus on only one constraint or bias at a time, without considering the relations among them.

Results revealed that participants were quite sensitive to maximum possible monetary returns but also showed robust deviations from maximizing as well (Erev et al., 2017). The evidence indicated that the deviations were due to “contradictory biases” that did not cancel each other out. This fits well with values-realizing theory, which claims that reasoning is guided by multiple values that are in tension with each other, and which are heterarchically related. The other constraints, besides economic value, that Erev et al. (2017) identified in their data were: (a) a tendency to make choices that would allow participants to avoid regret (i.e., would be easiest for them to justify to others and themselves; Connolly & Zeelenberg, 2002), (b) a pessimistic bias (i.e., a tendency to assume the worst possible outcome), (c) a tendency to weight factors relevant to expected outcomes equally (rather than assuming that future preferences would be weighted like past ones), (d) a preference for choosing the option that included the best possible outcome, and (e) a tendency to avoid excessively complicated possibilities (e.g., large payoff ranges, ambiguous outcomes). To a considerable extent these biases appear to reflect the importance of choices being jointly constrained by social solidarity, as well as clarity, coherence, comprehensiveness, and complexity (i.e., tension between best and worst possible outcomes).

Finding the common good

Decision making can be extraordinarily complicated even for people well-acquainted with the facts and dedicated to evidence-based reflection. In their discussion of risk and denial, Gorman and Gorman (2017) do a particularly good job of portraying the real-world difficulty of people trying to make good decisions in a complex world with multiple sources of information that may pull in different directions (see also Barzilai & Ka’adan, 2017). An example of the complexities of judgment is illustrated by a couple who had two close relatives they deeply admired, both medical professionals, strongly advise them against giving their first-born child the measles immunization sequence. Another relative, who was a scientist, was adamant that they should follow the standard protocol for childhood inoculations. After studying the various arguments, involving the risks to their child along with the countervailing risks to other children, they made a creative choice: they vaccinated, but adopted the longest possible time window for having the series of inoculations administered. This is an example of the kind of creative compromise that a values-realizing approach to reasoning encourages.

However, tragic outcomes can occur, especially if there is a focus on a single value, isolated from other values. If a value is isolated and treated as if it were a goal, other relevant values are likely to be shortchanged, undermining the possibility of realizing even the chosen value (Hodges & Baron, 1992). Safety alone can be dangerous, as parents in a town near me learned when they chose not to vaccinate their child. Their child got measles, affecting all his classmates: a week of school was lost for everyone. It did not endear the parents to the community, and their own child suffered. In the end, everyone was less safe.

The research reviewed earlier indicates that knowledge resistance is not primarily due to individuals with defective cognitive processes. Rather, it is the difficulty that all thinkers share in dealing with conflicting sources of information, and the tendency to avoid the resistance offered by disagreement and dialogue. It is extremely difficult to choose to trust other people when there are significant disagreements. Nevertheless, a values-realizing approach suggests that social solidarity and truth will require that trust be worked out, if there is to be a good chance of long-term success. If this is correct, it suggests scientists will need to establish social solidarity with all types of communities and cultures if they are to be effective in finding and sharing truths about problems and solutions. Thinkers of every kind need resistance from others, as well as agreement, if they are to find their way to truth, justice, and other values. Perhaps, in learning to trust and resist, researchers who think about thinking and do research on reasoning can be of help to others as well as themselves.