Ulrich Hoffrage

Together with his co-workers, Herbert Simon developed a research program on heuristics in scientific discovery. Simon saw heuristics as both practical tools for problem-solving, as well as descriptive models of how scientists and others solve problems. Building on Simon’s hybrid notion of heuristics as tools and theories, we extend Simon’s study of heuristics in scientific discovery to scientific justification. This article makes two types of contributions, inspirational-practical and conceptual-historical. First, we sketch out how one class of neo-Simonian heuristics, dubbed fast-and-frugal, and originally proposed as descriptive models of how minds make inferences, could enhance the transparency of scientific justification in model selection, and in so doing, the reproducibility of the conclusions drawn from empirical results. Furthermore, we explain how heuristics might serve meta-analyzing findings, aggregating quantitative indicators and qualitative considerations, identifying disconfirming and converging evidence, and examining the robustness of conclusions across different justification tools. Second, we embed our inspirational-practical proposal in a tour d’horizon on the conceptual associates with, and contrasts of, Simonian and neo-Simonian research on heuristics, journeying from early AI to a toolbox view on justification. With those inspirational-practical and conceptual-historical thrusts, we hope to address methodologists interested in ideas for fostering transparency in justification, as well as decision scientists interested in conceptual roots of research on fast-and-frugal heuristics. Finally, we outline what a neo-Simonian research program on heuristics as possible tools for, and descriptive models of, justification might look like.

How can people’s ability to make accurate estimations be boosted? Psychological research on the wisdom-of-the-inner-crowd suggests that people’s judgments improve when they use a simple consider-the-opposite-strategy, dubbed—inspired by Enlightenment philosopher Hegel—dialectical-bootstrapping: A person generates a first estimate (thesis), then rejects it and generates another one (anti-thesis), and finally integrates both (synthesis). Yet, the wisdom-of-the-inner-crowd-phenomenon comes with controversy concerning its measurement, robustness, and moderators. We (1) introduce a novel class of strategies to elicit the wisdom-of-the-inner-crowd. These strategies root in physics, where Nobel-laureate Enrico Fermi used back-of-the-envelope guesstimation, for instance, when assessing the explosive yield of the first tested nuclear bomb. Fermian strategies prescribe decomposing an estimation problem into subtasks, solving the subtasks separately, and ultimately integrating those solutions into a final estimate. In an experiment using a new task-environment, we find (2) that a similarity-based Fermian-strategy boosts the wisdom-of-the-inner-crowd even more than consider-the-opposite does, (3) that the provision of a memory aid differentially affects those two strategies’ performance, and (4) that data trimming matters. Moreover, and for the first time, we document (5) overprecision in wisdom-of-the-inner-crowd estimations. Finally, we (6) replicate previous results, including that the collective intelligence of two persons still outperforms asking oneself twice.

While Lee and Schwarz propose grounded procedures of separation as an explanation for physical cleansing in various domains (e.g., washing one's hands), we suggest that separation can also account for behavioral cleansing aimed at washing consciences and polishing reputations. We discuss this extension in terms of degrees of behavioral cleansing, motivations, and intentions behind cleansing, and social settings.

When shopping online, decision makers normally choose between alternatives by following a two-stage process; purchasing platforms support this process: After first reducing a set of alternatives by filtering on attributes, the platform’s users can, in a second step, evaluate the remaining alternatives in the consideration set simultaneously (i.e., jointly). In our research, we investigate how each of these two steps ― filtering and joint evaluation ― affect the success of a prosocial microlending platform. On such platforms, decision makers lend money interest-free to people in need of money, such as entrepreneurs from developing countries who want to run a small business. We hypothesize that while filtering in the first step can reduce the decision effort and provide guidance, which is often perceived as helpful in the domain of purchasing decisions, users of prosocial lending platforms may perceive such an interface as inappropriate and restrictive. Such a perception might decrease users’ choice satisfaction. We further build on evaluability theory to hypothesize that joint evaluation in the second step is a double-edged sword: While more than one alternative increases evaluability by allowing users to use alternatives as reference points, it can also easily create conflict between alternatives. Consequently, joint evaluation has the advantages of decreasing restrictiveness and increasing evaluability, which should leave users more satisfied with their choice. However, joint evaluation also has the disadvantage of prompting users to experience conflict between the alternatives and thereby decreases their willingness-to-contribute to the alternative they finally choose. We test our hypotheses in an incentivized lab experiment, using real prosocial lending decisions. Our findings suggest that offering typical attribute-based decision support, like filters, does not increase a platform’s success, and confirm that joint evaluation is a double-edged sword. We conclude that platforms have to trade off the decreased user choice satisfaction with their increased willingness-to-contribute.

Representing statistical information in terms of natural frequencies rather than probabilities improves performance in Bayesian inference tasks. This beneficial effect of natural frequencies has been demonstrated in a variety of applied domains such as medicine, law, and education. Yet all the research and applications so far have been limited to situations where one dichotomous cue is used to infer which of two hypotheses is true. Real-life applications, however, often involve situations (e.g., medical tests) where cues have more than one value, where more than two hypotheses are considered, or where more than one cue is available. In Study 1, we show that natural frequencies, compared to information stated in terms of probabilities, consistently increase the proportion of Bayesian inferences made by medical students in four conditions—three cue values, three hypotheses, two cues, or three cues—by an average of 37 percentage points. In Study 2, we show that teaching natural frequencies for simple tasks with one dichotomous cue and two hypotheses leads to a transfer of learning to complex tasks with three cue values and two cues, with a proportion of 40% and 81% correct inferences, respectively. Thus, natural frequencies facilitate Bayesian reasoning in a much broader class of situations than previously thought.

In Bayesian inference tasks, information about base rates as well as hit rate and false-alarm rate needs to be integrated according to Bayes’ rule after the result of a diagnostic test became known. Numerous studies have found that presenting information in a Bayesian inference task in terms of natural frequencies leads to better performance compared to variants with information presented in terms of probabilities or percentages. Natural frequencies are the tallies in a natural sample in which hit rate and false-alarm rate are not normalized with respect to base rates. The present research replicates the beneficial effect of natural frequencies with four tasks from the domain of management, and with management students as well as experienced executives as participants. The percentage of Bayesian responses was almost twice as high when information was presented in natural frequencies compared to a presentation in terms of percentages. In contrast to most tasks previously studied, the majority of numerical responses were lower than the Bayesian solutions. Having heard of Bayes’ rule prior to the study did not affect Bayesian performance. An implication of our work is that textbooks explaining Bayes’ rule should teach how to represent information in terms of natural frequencies instead of how to plug probabilities or percentages into a formula.