Anexo 6. Estrategia de imputación de casos perdidos
3.1.4. Comentarios sobre la incidencia del marco legislativo en el rol de
We turn now to the third component in the triangular scheme, sources. This thesis will argue that sources are inextricably linked to both rationality and
pragmatics. The link to pragmatics has already been explored above. This section will introduce the link between sources and rationality.
Within psychology, sources have been given the most extensive treatment in the psychology of persuasion and, in particular, in dual-route models. Although this
social psychological research is relevant for debates on rationality, it takes a purely descriptive approach. The persuasion literature will be surveyed in more detail in Chapter 6. Here, it suffices to note that information about sources – typically, a person making a claim - generally features in the peripheral, heuristic route (Briñol & Petty, 2009; Chaiken, 1980; Chaiken & Maheswaran, 1994; Petty & Briñol, 2008; Petty & Cacioppo, 1984, 1986). That is, sources influence persuasion primarily when the audience is unmotivated, unengaged or unskilled. They can influence persuasion indirectly in the central, analytic route through effects on processing and
metacognition. And, exceptionally, source information can be an explicit argument. Such sensitivity to sources seems intuitively reasonable, but the psychology of persuasion lacks an appropriate normative theory, the credibility of sources being determined, like the strength of arguments, in pre-tests (for further discussion, see Chapter 6).
A strikingly different approach is taken in a recent psychological theory, the Argumentative Theory of Reasoning (Mercier & Sperber, 2011, 2017). This theory places sources at the heart of reasoning, decision making, and judgement, in the sense that these cognitive abilities are seen as fundamentally geared towards interaction with others. On this view, reasoning (construed broadly to include decision making and judgement) is argumentative: it is not intended for deriving conclusions but rather for ‘[devising] and [evaluating] arguments intended to persuade’ (Mercier & Sperber, 2011, p. 57). Mercier and Sperber acknowledge evidence that, on an individual level, much of our reasoning seems to be biased or motivated (see, e.g., Kunda, 1990), but argue that this behaviour can be
argumentatively appropriate. For instance, it may seem undesirable for individual reasoners to generate evidence only in support of their belief, neglecting possible
falsification. But, Mercier and Sperber argue, such a confirmation bias primarily affects the production, not the evaluation of arguments, and allows an efficient division of labour, both sides producing as many arguments as possible and then, ideally, resolving their dispute together. Mercier and Sperber apply their theory to a wide range of classic fallacies. They argue, for instance, that people make choices that are easy to justify, regardless of their ultimate optimality, and argue that such reason-based choice can explain the attraction effect, disjunction effect, the sunk- costs fallacy, framing effects, and preference inversion (Mercier & Sperber, 2011). They also take a somewhat optimistic view of the data on group decision making, concluding that such data show that discussion improves reasoning (though, as we will below, see qualifications are due).
The claim that argumentation is social has precedent in Argumentation Theory. Research on argumentation asks what makes a good natural-language argument, whether people are good at natural-language argumentation, and whether people’s skills can be improved. Argumentation theory has taken a social turn, the focus being less on relationships between true and false propositions and more on the use of argument in reasoned dialogue (Toulmin, 1958; Walton, 2008). Since the social turn, argument strength is treated as relative to an audience (Perelman & Olbrechts-Tyteca, 1969; Toulmin, 1958; Toulmin, Rieke, & Janik, 1979). Argument strength can also vary across social contexts: for example, the same argument may be strong in a casual discussion but weak in a court of law. The content and contexts of argument can demand particular criteria for validity, criteria which are established by the participants in the argument (Toulmin, 1958; Toulmin et al., 1979).
More recent Argumentation Theory places even greater emphasis on social factors. This emphasis takes the form of rules intended to regulate the behaviour of participants. Take, for instance, the following rules from Pragma-Dialectics:
Rule 1 (Freedom Rule): Discussants may not prevent each other from advancing standpoints or from calling standpoints into question.
Rule 2 (Burden of Proof Rule): Discussants who advance a standpoint may not refuse to defend this standpoint when requested to do so
(Eemeren, Garssen, & Meuffels, 2009, pp. 21–22). These rules can be construed as social or ethical norms, their ethical character being clearer in the ‘fairness rules’ formulation of Christmann and colleagues (Christmann, Mischo, & Flender, 2000; Christmann, Mischo, & Groeben, 2000; Schreier, Groeben, & Christmann, 1995). Such rules have also been also put to work in explaining both qualitative (Eemeren & Houtlosser, 1999) and quantitative data (Christmann, Mischo, & Groeben, 2000; Eemeren et al., 2009; Schreier et al., 1995).
Sources figure especially strongly in a range of arguments which were classically deemed fallacious but which can, since the social turn, be considered good arguments. Historically, since sources were irrelevant to argument quality, arguments were deemed poor if they relied on appeal to features of a source (argumentum ad verecundiam) or attacking a source (argumentum ad hominem). However, properties of the source do seem to matter: intuitively relevant are its expertise, accuracy, and trustworthiness. Indeed, contemporary Argumentation Theory attempts to explain when source information is relevant and how it should be judged. Take, for instance, the following argumentation scheme and critical
questions for an argument from expertise from Walton, Reed, and Macagno (2008): Argument Scheme
Source E is an expert in subject domain S containing proposition A. E asserts that proposition A (in domain S) is true (false).
A may plausibly be taken to be true (false). Critical Questions
How credible is E as an expert source? Is E an expert in the field that A is in? What did E assert that implies A? Is E personally reliable as a source?
Is A consistent with what other experts assert? Is A based on evidence?
It is questionable whether such critical questions really do provide anything like a normative basis (Hahn & Hornikx, 2016). But, as this approach shows,
Argumentation Theory assumes that sources should be considered.
Related to both persuasion and argumentation is the philosophy of testimony. Testimony, in the philosophical sense, is information received from other people; the study of testimony has increasing impact on the psychology of rationality (see chapters in Zenker, 2012). Theories of testimony can provide both normative accounts and hypotheses to test. For testimony, the normative question is whether it is justifiable to revise one’s beliefs in response to testimony and, if so, how (see, e.g., Coady, 1992). Testimony has proved amenable to normative Bayesian models, which will be explored in Chapter 6 (Bovens & Hartmann, 2003; Olsson, 2011; Olsson & Vallinder, 2013). These models prescribe consideration of sources whenever information is received from other people; hence, they insist on a close relationship between rationality and sources. A growing literature tests whether these
models are descriptively accurate (e.g. Hahn, Harris, & Corner, 2009; Harris & Hahn, 2009; Harris, Hahn, Madsen, & Hsu, 2016).
Testimony overlaps with many tasks in judgment and decision making. This is perhaps most obviously the case in individual decision making by description, in which an individual makes a decision based on information provided to them by some source. Sources have made their way into the literature on individual judgement and decision making: recall the examples above in which experiments manipulated sources when exploring, for instance, the false-memory effect (Dodd & Bradshaw, 1980; Smith & Ellsworth, 1987) or ‘saying things that go without saying’ (Gruenfeld & Wyer, 1992). There is, of course, a more fundamental point:
experimental participants know that an experiment is designed by an experimenter to some end; they may factor this into their interpretation of materials, and into their decisions about whether to trust the information they receive (Hilton, 1995; Schwarz, 1996). There is scarce research on actual testimonial models in this context. A
notable exception is Bovens and Hartmann’s (2003) treatment of the conjunction fallacy, an analysis which shows that a sophisticated probabilistic reasoner could produce the typical response pattern (rating the conjunction higher than the conjuncts) under certain assumptions. This model has been tested by Jarvstad and Hahn (2011), who found no evidence to support the model’s assumptions. This issue will resurface in Chapter 4.
Testimony also overlaps with group judgement and decision making. Although there are testimonial models of receiving information from multiple sources (e.g. Bovens & Hartmann, 2003), these models do not seem to have had an impact on the group decision-making literature. There is considerable potential for useful dialogue between the two disciplines. A key question is how individual and
group performance are related: under what conditions does group performance outperform individual performance (for a review, see Gigone & Hastie, 1997; Kerr & Tindale, 2004)? The main focus has been on interaction: for instance, studies have had groups choose and nominate their best member (the dictator method); or freely debate to reach consensus (the consensus method); or consider the group mean, think of reasons why it might be too high or low, and then decide on an answer (the
dialectic method); or just view all judgements, and make revisions in rounds, without discussion, until consensus is achieved (the Delphi method) (eg. Sniezek, 1989). The crucial measure is how the group performs relative to some baseline, often the performance of the most successful member (Kerr & Tindale, 2004).
Groups vary from the baseline by making process gains (they outperform their best member) or process losses (they are outperformed by their best member). These deviations from the baseline give insight into rationality. Take, first, process losses. There are numerous examples in the literature. When there is discussion, one party can dominate, trains of thought can be prevented from arising or, if they do arise, can be derailed (Kerr & Tindale, 2004). Other common deficits are
demotivation and social loafing (Kerr & Tindale, 2004). Groups can also exchange information sub-optimally. For instance, in the ‘hidden profile method’, some information is distributed to all participants (shared information); other information is distributed only to some (unshared information) (Stasser & Titus, 1985, 2003). Participants tend to overweight shared information, especially under time pressure; they prefer to present and receive shared information; and they judge someone who gives shared information as more competent and credible (for discussion, see Kerr & Tindale, 2004). There are, nevertheless, cases of process gains. When groups are cohesive, share strong productivity norms, and have a shared understanding of the
task, they show high motivation, and can perform strikingly well (Mellers et al., 2014). Such factors were exploited in the recent Good Judgment Project, in which groups who discussed forecasts freely outperformed both individuals and groups who only saw other members’ judgments (see, e.g., Atanasov et al., 2015; Mellers et al., 2014; Mellers, Stone, Atanasov, et al., 2015; Mellers, Stone, Murray, et al., 2015; Tetlock, Mellers, Rohrbaugh, & Chen, 2014).
Such experiments offer rich testimonial contexts and ample opportunities for models of complex testimony with multiple sources (see Bovens & Hartmann, 2003, and for discussion Hahn, Harris, & Corner, 2016). These rewarding prospects argue for further research to explore how people interact and how they process the
information they receive from sources. Such fine-grained manipulations are unfeasible in such large-scale studies as the Good Judgment Project. There is, however, suggestive evidence already. Participants seem to consider sources expert if they are loquacious, influence other group members using reason, and are
confident or dominant (Kerr & Tindale, 2004; for the empirical studies, see Littlepage & Mueller, 1997; Littlepage, Schmidt, Whisler, & Frost, 1995). Much remains to be done to explore the rationality of both these reliability judgements and their subsequent use in weighting evidence.
Although such group decision making is an extremely valuable topic of research, this thesis will focus on individual judgement, reasoning, and decision making by description. This focus does not imply that this is the more important level. Rather, these individual judgements mesh better with natural-language pragmatics, which has focused, thus far, on dyadic or small-scale interactions. This focus also allows us to explore basic mechanisms which can later be extended to
group contexts. As we will see in Chapters 2-4 and 6, there is much to be done at this lower level.