Uncertainty Communciation

This project studies how the communication of scientific uncertainty shapes both (i) how people evaluate scientific evidence and (ii) how they act on that evidence when making a constrained support decision. In many real-world settings—policy, philanthropy, program funding, and public decision-making—people must interpret empirical findings that are uncertain (e.g., confidence intervals, sampling variation, imperfect measurement) and then decide what to support. Yet uncertainty can be communicated in different ways: sometimes it is explicitly disclosed (e.g., with ranges or error bars), and sometimes it is omitted (e.g., point estimates only). Likewise, results can be communicated in text-only formats or accompanied by visuals, which may improve comprehension but can also change how credible, transparent, or persuasive the evidence seems. The central purpose of this study is to identify which communication formats help people interpret uncertainty appropriately without undermining trust, and whether these communication choices meaningfully affect evidence-based support for competing programs. The project contributes to ongoing debates in science communication and evidence-informed policymaking about whether communicating uncertainty improves transparency and understanding or instead creates doubt, confusion, or scepticism. Study context and motivation Decision-makers frequently face evidence that does not cleanly point to a single dominant choice. Instead, they must compare options where one program appears to have a higher estimated impact but also greater uncertainty, while another appears more precisely estimated but with a smaller estimated effect. This type of trade-off is common in evaluations and applied social science, and it mirrors the kinds of judgments made when allocating limited resources across interventions. The study is motivated by the practical question: How should researchers and institutions communicate uncertain findings so that audiences can weigh evidence accurately and make decisions that are aligned with the information provided? Overview of study design The project is implemented as a survey experiment with a 2 × 2 factorial design. Respondents are randomly assigned to view evidence about two education interventions in one of four communication formats created by crossing: Uncertainty disclosure Explicit uncertainty disclosed: Results are presented with a point estimate and a numeric uncertainty range (and when visual, error bars). No uncertainty disclosed: Results are presented as point estimates only (and when visual, dots without uncertainty intervals). Modality Text-only: Evidence is conveyed using written descriptions and numerical results. Visual + text: Evidence is conveyed using written descriptions and accompanying plots (a single-intervention plot at first exposure and a combined comparison plot at the decision stage). Within each format, the study also includes counterbalancing to prevent results from being driven by order effects or label preferences: the order in which interventions are shown is varied, and which intervention label is paired with the larger estimated effect is randomized. In the core experimental conditions, evidence is constructed to present an ambiguous trade-off: one option has a larger estimated effect but more uncertainty, and the other has a smaller estimated effect but greater precision. This structure is intentional because it reflects realistic decision environments where uncertainty matters and where the evidence does not provide a single obviously superior choice. Measures and outcomes Before any experimental manipulation, respondents complete a common pre-treatment module measuring: baseline trust (generalized trust and trust in institutions such as researchers, government, NGOs, news media, social media, and personal networks), prior beliefs relevant to the domain (e.g., beliefs about education quality and whether more should be done), past prosocial behaviour (e.g., donating, volunteering, advocacy-related activities), baseline beliefs about the effectiveness of each intervention type. After viewing the first evidence vignette, respondents report their perceptions of the evidence and its source using a standardized battery capturing: trust in the researchers, perceived accuracy and precision of the findings, perceived transparency and clarity of communication, perceived sufficiency of information, perceived replicability and whether key details seem omitted. These measures capture the study’s first outcome domain: perceived credibility and communication quality. Next, respondents view evidence about the second program and then make a behavioural allocation decision: they allocate a fixed total budget (500 units) across the two programs. This forced-sum allocation is used as a proxy for policy or funding support under constrained resources. Respondents also explain their decision in open text and answer interpretation/comprehension checks about which program had higher estimated impact and which was more uncertain. obustness and validation components In addition to the core factorial conditions, the design includes validation/robustness arms that are not the primary focus but strengthen interpretation. One set presents an obvious dominance case in which one option clearly outperforms the other (higher estimated effect with non-overlapping uncertainty), allowing assessment of whether participants respond sensibly when evidence is clear and whether visuals improve decision quality in a straightforward setting. Another arm presents a no-evidence baseline in which participants allocate funds without seeing any results, allowing estimation of baseline splitting behaviour and any residual label preferences. Expected outcomes and contributions The study is designed to produce causal evidence on: The effect of explicit uncertainty disclosure on trust, perceived credibility, transparency, clarity, and perceived sufficiency of evidence. The effect of visualization on comprehension-related perceptions (clarity, precision, confidence in understanding) and whether visuals amplify or attenuate any effects of uncertainty disclosure. Behavioural consequences of uncertainty communication—specifically, whether communication formats change how respondents allocate support across options when faced with a realistic trade-off between higher-but-uncertain and lower-but-precise evidence. Mechanisms and boundary conditions, assessed through interpretation checks and pre-registered heterogeneity analyses using baseline trust and prior beliefs. Overall, the project aims to inform best practices for communicating uncertain empirical findings in settings where decisions depend on evidence. It generates practical guidance for researchers, evaluators, and institutions about when uncertainty should be made explicit, whether visual displays help audiences use evidence appropriately, and how communication choices influence both perceived credibility and real allocation behaviour.