The perceived functions of linguistic risk quantifiers and their effect on risk, negativity perception and decision making

Subjective Probability Increases Across Communication Chains: Introducing the Probability Escalation Effect

A severity effect has previously been documented, whereby numerical translations of verbal probability expressions are higher for severe outcomes than for non-severe outcomes. Recent work has additionally shown the same effect in the opposite direction (translating numerical probabilities into words). Here, we aimed to test whether these effects lead to an escalation of subjective probabilities across a communication chain. In four 'communication chain' studies, participants at each communication stage either translated a verbal probability expression into a number, or a number into a verbal expression (where the probability to be translated was yoked to a previous participant). Across these four studies, we found a general Probability Escalation Effect, whereby subjective probabilities increased with subsequent communications for severe, non-severe and positive events. Having ruled out some alternative explanations, we propose that the most likely explanation is in terms of communications directing attention towards an event's occurrence. Probability estimates of focal outcomes increase across communication stages.

A comparison of human and GPT-4 use of probabilistic phrases in a coordination game

English speakers use probabilistic phrases such as likely to communicate information about the probability or likelihood of events. Communication is successful to the extent that the listener grasps what the speaker means to convey and, if communication is successful, individuals can potentially coordinate their actions based on shared knowledge about uncertainty. We first assessed human ability to estimate the probability and the ambiguity (imprecision) of twenty-three probabilistic phrases in a coordination game in two different contexts, investment advice and medical advice. We then had GPT-4 (OpenAI), a Large Language Model, complete the same tasks as the human participants. We found that GPT-4's estimates of probability both in the Investment and Medical Contexts were as close or closer to that of the human participants as the human participants' estimates were to one another. However, further analyses of residuals disclosed small but significant differences between human and GPT-4 performance. Human probability estimates were compressed relative to those of GPT-4. Estimates of probability for both the human participants and GPT-4 were little affected by context. We propose that evaluation methods based on coordination games provide a systematic way to assess what GPT-4 and similar programs can and cannot do.

Do people really prefer verbal probabilities?

When people communicate uncertainty, do they prefer to use words (e.g., "a chance", "possible") or numbers (e.g., "20%", "a 1 in 2 chance")? To answer this question, past research drew from a range of methodologies, yet failed to provide a clear-cut answer. Building on a review of existing methodologies, theoretical accounts and empirical findings, we tested the hypothesis that the preference for a particular format is driven by the variant of uncertainty that people experience. We expected that epistemic uncertainty would be more often communicated in words, whereas distributional uncertainty would be more often communicated in numbers; for the dispositional uncertainty, we expected that an individual's disposition would be more often communicated in words, whereas dispositions from the world would be more often communicated numerically. In three experiments (one oral, two written), participants communicated their uncertainty regarding two outcomes per variants of uncertainty: epistemic, dispositional and distributional. Overall, participants communicated their uncertainty more often in words, but this preference depended on the variants of uncertainty. Participants conveyed their epistemic and dispositional uncertainties more often in words and their distributional uncertainty in numbers (Experiments 1 and 2) but this effect was greatly reduced when the precision of uncertainty was held constant (Experiment 3), pointing out the key role of uncertainty vagueness. We have reviewed the implications of our findings for the existing accounts of format preferences.

Politeness and the communication of uncertainty

Ambiguity in language derives, in part, from the multiple motivations that underlie the choice to use any particular expression. The use of some lexical items, such as probability expressions and scalar terms, can be motivated by a desire to communicate uncertainty as well as a desire to be polite (i.e., manage face). Research has demonstrated that the interpretation of these items can be influenced by the existence of a potential politeness motive. In general, communications about negative events, relative to positive events, result in higher likelihood estimates whenever politeness can be discerned as a potential motive. With few exceptions, however, this research has focused only on the hearer. In the present research we focused on the dyad and examined whether speakers vary their messages as a function of politeness, and the effect that this has on subsequent judgments made by a recipient. In two experiments we presented participants with situations that varied in terms of face-threat and asked them how they would communicate potentially threatening information. Both experiments included a second set of participants who read these utterances and provided judgments as to the degree of uncertainty conveyed by the utterance. In both experiments, messages in the face-threatening condition conveyed greater uncertainty than messages in the non-face-threatening condition, and the probability estimates made by the second set of participants varied as a function of conveyed uncertainty. This research demonstrates that when examining speakers and hearers together, severe events may be judged less likely (rather than more likely), because speakers tend to hedge the certainty with which they communicate the information.

A direct and comprehensive test of two postulates of politeness theory applied to uncertainty communication

Applied to uncertainty communication, politeness theory postulates that when announcing bad news (1) speakers may intend not only to inform, but also to manage (e.g., save) the hearers’ or speakers’ own faces (i.e., face-managing intentions), and (2) speakers may perform face-managing intentions by altering the explicitly communicated probability. Previous research has assumed these two core postulates when explaining various reasoning and judgment phenomena in hearers, but has failed to test them empirically in a comprehensive and direct way: jointly in relation to speakers. To provide this critical evidence, we asked subjects to communicate a predefined numerical probability of two negative outcomes, using a verbal probability scale. Subjects reported their communication intentions afterwards. In line with the first politeness theory postulate, speakers intended not only to be informative but also to tactfully announce bad news or to avoid being blamed in case they made inaccurate (too low or too high) prediction. In line with the second politeness theory postulate, speakers altered their explicitly communicated probability more often and more substantially when adopting face-managing intentions than when adopting informative intentions. We discuss how this evidence corroborates the politeness theory and validates the previous research that focused on hearers.

Do people really say it is "likely" when they believe it is only "possible"? Effect of politeness on risk communication

Politeness theory posits that speakers can use verbal probabilities (e.g., there is a chance, it is likely) to hedge bad news. So far, only indirect evidence supports that claim: From the hearer's standpoint, verbal probabilities are interpreted either as plain likelihood-communication devices or as face-management devices, resulting in different risk perceptions. The present research aims to test more directly the postulate of politeness theory by focusing on the effects of speakers' intentions on risk communication. In three experiments, participants communicated a probability by choosing an expression from a list of verbal probabilities. Results consistently showed that polite speakers communicated a different risk magnitude than informative speakers. Further findings indicate that the effect of the speakers' intention depends on the valence of the uncertain outcome. The theoretical and applied implications of these findings are discussed.

Top scores are possible, bottom scores are certain (and middle scores are not worth mentioning): A pragmatic view of verbal probabilities

In most previous studies of verbal probabilities, participants are asked to translate expressions such as possible and not certain into numeric probability values. This probabilistic translation approach can be contrasted with a novel which-outcome (WO) approach that focuses on the outcomes that people naturally associate with probability terms. The WO approach has revealed that, when given bell-shaped distributions of quantitative outcomes, people tend to associate certainty with minimum (unlikely) outcome magnitudes and possibility with (unlikely) maximal ones. The purpose of the present paper is to test the factors that foster these effects and the conditions in which they apply. Experiment 1 showed that the association of probability term and outcome was related to the association of scalar modifiers (i.e., it is certain that the battery will last at least…, it is possible that the battery will last up to…). Further, we tested whether this pattern was dependent on the frequency (e.g., increasing vs. decreasing distribution) or the nature of the outcomes presented (i.e., categorical vs. continuous). Results showed that despite being slightly affected by the shape of the distribution, participants continue to prefer to associate possible with maximum outcomes and certain with minimum outcomes. The final experiment provided a boundary condition to the effect, showing that it applies to verbal but not numerical probabilities.

To what extent do politeness expectations shape risk perception? Even numerical probabilities are under their spell!

Politeness theory posits that uncertainty quantifiers can be interpreted as hedging strategies and thus be perceived as communicating greater certainty than when they are interpreted as likelihood-communication devices. This has only been tested with verbal probabilities (e.g., it is possible). The present paper aims to test whether numerical probabilities can also be interpreted as face-management devices and to investigate the effect of such an interpretation on risk perception. Four experiments focused on the effect of interpretations of numerical probabilities in negative outcome predictions on risk perception (e.g., there is a 50% probability that your stocks will lose their value). Politeness expectation was manipulated by the personality of the speaker (i.e., blunt vs. tactful, Experiments 1 and 2) and according to the conversational partners' need for politeness (Experiments 3 and 4). Results show that numerical probabilities, like verbal ones, were interpreted as likelihood-communication or face-management devices and that the two interpretations led to different risk perceptions. Findings were replicated with different formats, such as percentage (e.g., 50%) and chance ratio (e.g., 1 chance in 3) and with different degrees of certainty (e.g., .30, .50 and .70). Theoretical and practical implications relevant to risk communication are presented and discussed.