Could evaluation mode nudge people's health-related decisions? Extending the general evaluability theory to risk communication

• Main Author: Xu, Yihan
• Other Authors: Charles T. Salmon
• Format: Thesis-Doctor of Philosophy
• Language: English
• Published: Nanyang Technological University 2020
• Subjects: Social sciences::Communication::Communication theories and models; Social sciences::Psychology::Experimental psychology
• Online Access: https://hdl.handle.net/10356/142967
• Institution: Nanyang Technological University

Many important decisions people make are related to health. However, people often show inconsistent preferences, even for life-or-death matters. Preference inconsistency happens because people’s preferences are often constructed on the spot — not as stable as normative decision theories assume. Preference reversals can be induced by many factors, and notably, by evaluation mode —the focus of this dissertation. Evaluation mode refers to the condition under which options are presented and evaluated. People engage in single evaluation (SE) mode if they evaluate one single option in isolation, and they engage in joint evaluation (JE) mode if they evaluate two (or more) options simultaneously, and their preferences may reverse between SE and JE. As the evaluability hypothesis and general evaluability theory (GET) predict, when people evaluate two options that involve trading off between an easy-to-evaluate attribute (e.g. condition) and a hard-to-evaluate attribute (e.g. number of entries), their preferences for the two options may reverse between SE and JE. In JE, the hard-to-evaluate attribute becomes easier to evaluate and thus has more impact. For example, students prefer the music dictionary with a neat cover and 10,000 entries in SE, but they prefer the one with a torn cover and 20,000 entries in JE, presumably it is easier to evaluate the condition (new vs. old) than the usefulness of the dictionaries. Whereas previous research on preference reversals between SE and JE has accumulated mostly in the domain of consumer behavior and managerial decision-making, this dissertation demonstrates SE/JE preference reversals in health-related decisions that involve risks. The extent to which laypeople can make good health-related decisions is often dependent on how reasonably they perceive the riskiness of the available options. Risk is a multifaceted concept with several important dimensions that tap into both emotion and cognition — while emotion-laden attributes such as severity and pain intensity may be easy to evaluate, numerical attributes such as likelihood and prevalence are often hard to evaluate. Therefore, it is often challenging for laypeople to evaluate the risks holistically and retain their evaluations under different evaluation modes. It is thus hypothesized that when laypeople face two options that involve a trade-off between an easy-to-evaluate risk attribute and a hard-to-evaluate risk attribute, their risk preferences will shift (even reverse) between SE and JE. More specifically, in SE, laypeople will prefer the option superior in the easy-to-evaluate attribute; in JE, they will prefer the option superior in the hard-to-evaluate attribute. This dissertation tests this main hypothesis and its derivative hypotheses with three progressively designed experiments. Study 1a examines laypeople’s willingness to accept a potentially risky business travel assignment to a foreign country, where they trade off severity of loss (easy-to-evaluate) and likelihood of loss (hard-to-evaluate). The results show moderate preference shifts between SE and JE, though the direction of JE/SE prediction contradicts the prediction. Study 1b examines laypeople’s perceived price fairness of a costly magic pill that offers a lifetime exemption of two potentially severe health hazards (stroke vs. stomach cancer), where they trade off intensity of loss (easy-to-evaluate) and magnitude of loss (hard-to-evaluate). The results conceptually replicate the findings from study 1a. Contrary to what the GET-derived hypothesis predicts, the results of study 1 show that the easy-to-evaluate risk attribute, rather than being crowded out by the hard-to-evaluate one, becomes even more impactful in JE. A risky option’s relative inferiority in the easy-to-evaluate risk attribute heightens, rather than abates, in JE. Study 2 aims to clarify the unexpected findings from study 1, expanding the scope of study 1 by investigating the boundary conditions of GET-derived hypothesis. It examines how laypeople trade off a risk attribute and a non-risk attribute when attributes’ justifiability (how relevant an attribute is for the decision-making task) vary. Study 2 investigates how the relative justifiability of costs affects laypeople’s evaluations of two flu vaccines, where they trade off costs (easy-to-evaluate) and chance of side effects (hard-to-evaluate). The results show that consistent with the hypothesis, preferences reverse between SE and JE, but to a greater extent when the costs is relatively more justifiable than when it is less so. The results also suggest that the pattern of hypothesized preference reversals is contingent upon an attribute’s relative justifiability. The findings from the two studies show that evaluation mode could nudge people’s health-related decisions. When laypeople have to deal with two options that are risky in a more or less intuitive sense, they generally prefer to avoid the more intuitively risky one in SE, but prefer to avoid the less intuitively risky one in JE. However, when the options are too risky overall, they may dread both options more when comparing them together than examining them one by one. How their preferences change from SE to JE is also somewhat contingent on the relative justifiability of the attributes in questions. In general, laypeople’s preferences are more malleable in SE, but more stable in JE, indicating that JE could serve as a buffer for contextual cues. Overall, this dissertation extends the GET framework to the context of health-related decision-making that involves risks and identifies one boundary condition for it. It also offers valuable insights for designing effective risk communication for risk communicators.