Education Increases Solar Radiation Modification Literacy but Reinforces Caution: Evidence from a Pre–Post University Study

Abstract

Solar Radiation Modification (SRM) is increasingly discussed as a potential supplement to climate-change mitigation, yet public and stakeholder judgments remain sensitive to knowledge, framing, and perceived risks. We examined how a structured university classroom module on SRM reshaped student perceptions using a matched pre–post survey design. Participants were students enrolled in an English-taught global climate change course (N = 106); 103 students provided valid matched responses after applying pre-specified exclusion rules. Self-rated SRM knowledge increased substantially after the module (mean change +0.47 on a 1–3 scale; Wilcoxon signed-rank p (Holm-adjusted) < 1 × 10⁻⁷; Cohen’s dz = 0.67). Support for SRM research remained moderately positive but did not increase (pre mean 3.76 to post mean 3.54 on a 1–5 scale). In contrast, support for stratospheric aerosol injection (SAI) deployment declined (pre mean 3.42 to post mean 2.95; p (Holm-adjusted) = 0.0084; dz = −0.33), and preferences shifted away from prioritizing climate intervention toward low-carbon development (mean change −0.68 on a 1–5 priority scale; p (Holm-adjusted) = 0.0001; dz = −0.45). Post-lecture models indicated that perceived benefits versus risks was the most consistent correlate of support across outcomes. Open-ended responses most frequently emphasized feasibility, unintended consequences, governance, and moral hazard. Overall, students largely endorsed SRM research as valuable while becoming more cautious about deployment and political prioritization, suggesting that balanced, structured instruction can sharpen sensitivity to evidence, uncertainty, and potential trade-offs that students also weighed in the survey.

1. Introduction

Climate change is already generating widespread and intensifying risks across natural and human systems, while the pace and scale of mitigation and adaptation remain uneven and, in many settings, insufficient to prevent severe impacts [1,2,3]. In this context, “climate intervention” has increasingly entered scientific and policy discourse as an umbrella term for approaches that either remove CO₂ (carbon dioxide removal, CDR), reduce incoming shortwave radiation, or otherwise alter Earth’s radiative balance (solar radiation modification, SRM) [4,5,6]. SRM—often discussed through proposals such as stratospheric aerosol injection or marine cloud brightening—could in principle lower global mean temperature on relatively short timescales, but it does not address the root cause of climate change (elevated greenhouse-gas concentrations) and introduces substantial uncertainties, trade-offs, and governance challenges [3,4,7].

The modern SRM debate is frequently traced to renewed scientific attention in the early 2000s, including the influential discussion of stratospheric sulfate injection as a “policy dilemma” response to the conflicting effects of cleaning up air pollution and keeping warming under control [6]. Yet, scholars have emphasized that SRM proposals exist within a contested socio-technical landscape shaped not only by physical climate responses, but also by ethical concerns, political legitimacy, justice, and geopolitical risk [8,9,10,11,12,13]. Authoritative assessments have repeatedly concluded that SRM cannot be considered a “silver bullet,” that evidence about regional impacts and unintended consequences remains limited, and that any SRM research agenda must be accompanied by robust, transparent governance [4,5,7,8].

Because SRM would entail deliberate, potentially transboundary climate influence, decisions about research trajectories and any prospective deployment raise legitimacy questions that are inseparable from public understanding and societal deliberation [14,15]. However, empirical studies consistently show that public awareness of geoengineering is low; attitudes are often “soft,” contingent, and sensitive to framing, trust, perceived naturalness, and perceptions of risk–benefit distribution [16,17,18,19]. In addition, recent syntheses argue that perceptions research has been disproportionately concentrated in Global North settings and has under-represented youth perspectives, despite the intergenerational stakes of SRM governance [19,20].

Universities represent a strategically important site for studying how individuals form and revise judgments about SRM. Undergraduate students are future scientists, engineers, policymakers, business leaders, and educators who will help shape institutional and societal responses to climate risks and emerging technologies [21,22]. Meanwhile, the climate change education (CCE) literature indicates that well-designed educational interventions can improve climate-related knowledge and, under some conditions, shift attitudes and behavioral intentions—although effects vary by pedagogy, context, and outcome domain [23,24]. Yet SRM presents distinctive instructional challenges: it is a “wicked” and value-laden topic with deep uncertainty, ethically fraught distributional implications, and potential for moral hazard or politicization [9,25,26]. Understanding whether and how formal instruction changes students’ knowledge structures, perceived legitimacy, and governance preferences is therefore not only an educational question, but also part of anticipatory governance for emerging climate interventions [27].

This study addresses that gap by examining cognitive and evaluative changes in undergraduate views of solar radiation modification (SRM) following a structured classroom teaching module embedded in a university-wide climate change course taught in English at Zhejiang University. In this paper, we use “learning” in an educational-evaluation sense—capturing changes in students’ self-rated knowledge, conceptual literacy, and evaluative perceptions after instruction—rather than claiming objective mastery of SRM facts. A key pedagogical feature is perspective-taking through regional role assignment: students are organized into six groups—East Asia, Southeast Asia, South America, North America, Europe, and Africa—to encourage deliberation about SRM’s uneven risks, benefits, and governance dilemmas from heterogeneous regional standpoints. By combining a matched pre-post survey with a learning design explicitly oriented toward distributional justice and geopolitical asymmetries, the study generates evidence on how climate intervention education can shape (i) SRM knowledge and perceived scientific feasibility; (ii) perceived risks and benefits; (iii) ethical and justice reasoning; and (iv) preferences for research governance and deployment constraints. In doing so, it contributes to the emerging intersection of SRM governance scholarship and higher-education climate pedagogy, with empirical evidence from an East Asian elite-university context that remains underrepresented in existing perceptions research [28,29].

2. Literature Review

2.1. SRM in the Climate Intervention Landscape: Promise, Limits, and Uncertainty

Solar radiation modification refers to a family of approaches intended to cool the Earth by increasing planetary albedo or otherwise reducing absorbed solar energy [3,4,6]. Compared with mitigation and many forms of CDR, SRM is sometimes characterized as “fast” and potentially less costly in direct implementation terms, but it is also “imperfect”: model-based evidence suggests that SRM could reduce some hazards associated with global mean warming while leaving residual regional climate change, shifting precipitation patterns, and creating novel risks stemming from design choices and governance arrangements [4,7]. Importantly, SRM does not resolve ocean acidification nor does it address existing carbon emissions. One significant tradeoff from this is dubbed the termination shock, which is the risk that if SRM is deployed but then abruptly terminated without any reductions of atmospheric greenhouse gases, then temperatures will rebound quickly and at a rate that is potentially extremely damaging to society and the biosphere [30,31].

Foundational SRM discussions illustrate the core policy dilemma: if mitigation lags and climate impacts escalate, societies may face pressure to consider additional interventions—even while acknowledging that these interventions are themselves uncertain and ethically contentious [32,33]. Subsequent assessments emphasize that SRM research remains fragmented, knowledge gaps are substantial, and any move toward outdoor experimentation or operational planning would raise governance questions about transparency, liability, consent, and international coordination [4,6,7]. This combination—potential risk reduction alongside profound uncertainty and moral controversy—makes SRM a paradigmatic “socioscientific” issue whose public meaning is likely to be shaped as much by values and political imaginaries as by physical science alone [14,15,25,26,34].

2.2. Governance, Legitimacy, and Justice as Inseparable Dimensions of SRM Perceptions

A central theme in SRM scholarship is that governance is not a downstream add-on; rather, it is constitutive of whether SRM research can be considered legitimate and socially responsible [13,34,35]. Proposals for SRM governance frequently invoke principles associated with “responsible innovation,” including anticipation of harms, reflexivity about assumptions and values, inclusion of diverse publics, and responsiveness to societal concerns [34]. Within SRM specifically, legitimacy concerns are amplified by the technology’s transboundary nature and the prospect of uneven climatic, ecological, and political consequences across regions, as well as potential conflicts or tradeoffs with sustainability goals and development [13,34,35,36,37].

Ethical debates also foreground intergenerational and distributive justice: who bears risks, who gains benefits, who decides, and under what conditions [9,10]. These issues intersect with practical governance challenges such as “termination shock” risks, potential weaponization narratives, and geopolitical conflict imaginaries [30,38,39]. For education research, this matters because perceptions of SRM are not merely beliefs about technical effectiveness; they are often judgments about fairness, institutional trust, and the moral acceptability of intentionally “messing with nature” [15,40,41]. Such judgments are likely to be especially salient when students reason from different regional vantage points, because regional role assignment makes distributional questions cognitively and emotionally concrete.

A further governance-related concern is “moral hazard”: the possibility that imagining SRM as an “insurance policy” could reduce perceived urgency for mitigation or shift political will away from emissions reductions [41,42,43]. Experimental work indicates that the relationship between SRM information and mitigation attitudes is not straightforward; responses can vary because of prior climate skepticism, trust, and framing, implying that education may not simply “increase support” but could also increase ambivalence, conditional acceptance, or stronger demands for constraints and oversight [41,42,43]. This underscores the need for empirical designs that measure how and why attitudes change during instructional exposure and learning processes, rather than assuming linear “knowledge to acceptance” trajectories.

2.3. What Shapes Public and Stakeholder Perceptions of SRM?

A robust body of perceptions research suggests several consistent patterns. Across studies, prior familiarity with geoengineering is generally limited, meaning early impressions can be highly sensitive to the first frames and analogies people encounter [14,15,16,17,40,44]. Survey evidence has shown that support for SRM tends to decline, and uncertainty tends to increase, when respondents move from abstract endorsement to more concrete deployment scenarios; acceptance is often conditional on governance safeguards, transparency, and perceived necessity [14,16,17,19]. Beyond familiarity and scenario specificity, perceived “naturalness,” perceived controllability, and moral intuitions about human–nature relationships can be as influential as technical cost–benefit reasoning [16,44].

In addition, deliberative and qualitative studies highlight that the public often contextualizes SRM within broader narratives of political failure, inequality, and distrust, rather than evaluating it as an isolated technology [15,44,45,46]. For example, analyses of public discourse describe SRM as a “global social experiment” that may intensify concerns about democratic governance and geopolitical conflict [15]. Deliberative engagement around the Stratospheric Particle Injection for Climate Engineering (SPICE) project demonstrated willingness to entertain limited field testing under strict conditions, while expressing discomfort about large-scale deployment and emphasizing the importance of openness and accountability [45].

Converging evidence also points to the central role of framing. Describing geoengineering by analogy to natural processes can increase support, indicating that cognitive heuristics and interpretive frames strongly shape acceptance—especially under low prior knowledge [20,43,47]. Related work shows that discussions can polarize around whether SRM is an unacceptable violation of nature or a pragmatic harm-reduction measure under emergency conditions [44]. These findings motivate education designs that explicitly surface and interrogate frames rather than presenting SRM as a neutral technical option.

Finally, cross-cultural evidence indicates that SRM perceptions vary across regions and policy scenarios. Recent survey research across multiple Asia–Pacific countries finds scenario-dependent attitudes and highlights the important of governance and risk perceptions in shaping acceptance [29]. Meanwhile, recent meta-level analyses of the “global perspectives” literature argue that many studies still center around Global North publics and underrepresent youth, raising justice concerns about whose voices are treated as epistemically and politically consequential for SRM futures [19,20]. Together, these insights support the pedagogical rationale of assigning students to regional groups and asking them to reason from differentiated climate vulnerability and geopolitical positions—an approach aligned with calls for more inclusive and justice-oriented SRM deliberation.

2.4. From Perceptions to Learning: Why Education Is Not a Simple “Information Deficit” Fix

Although SRM perceptions are often studied through surveys and deliberative forums, fewer studies have examined changes in judgments through formal instruction within higher education settings. This matters because universities are among the few institutions capable of providing sustained, scaffolded engagement with complex climate technologies, integrating physical science with ethics, governance, and regional political economy [21,22]. However, climate change education research cautions against assuming that providing more facts automatically yields more “pro-climate” attitudes or uniform behavioral change. Meta-analytic and systematic-review evidence indicates that climate change education can improve climate literacy (knowledge, attitudes, behavioral intentions), but effects are heterogeneous and moderated by intervention design, duration, learner characteristics, and outcome measures [23,24]. Parallel calls in solar geoengineering scholarship emphasize that social-science research—including education- and deliberation-based empirical designs—is essential for understanding how information, framing, and governance contexts shape public judgments [48].

Education theory provides useful lenses for SRM learning research. From a transformative learning perspective, confronting dilemmas that unsettle existing assumptions can catalyze critical reflection and perspective change, especially when learners engage dialogically with alternative viewpoints [49]. Conceptual change accounts similarly suggest that learners revise mental models when existing conceptions are challenged, intelligible alternatives are offered, and new explanations are perceived as plausible and fruitful [50]. SRM instruction plausibly triggers conceptual change because it confronts students with counterintuitive ideas, deep uncertainty, and competing policy logics. Meanwhile, socioscientific issues (SSI) frameworks argue that learning should include informal reasoning, evaluation of evidence, moral and ethical deliberation, and argumentation under uncertainty—precisely the competencies demanded by SRM governance debates [51,52,53].

For SRM, a key pedagogical implication is that “learning outcomes” should be conceptualized multidimensionally. Beyond factual knowledge, relevant outcomes include risk–benefit reasoning, ethical judgment, perceived legitimacy, trust in governance, and the ability to articulate conditional policy positions [14,15,16,17,45]. SRM is often discussed as a response to possible mitigation shortfalls. Instruction may therefore reshape students’ perceptions of mitigation urgency and policy priority—for example, by reinforcing the need for emissions reductions (SRM as imperfect and temporary) or by raising moral-hazard concerns if SRM is interpreted as a substitute [19,42,43].

2.5. Research Gap and Positioning of the Present Study

Syntheses of SRM perceptions research emphasize that early attitudes are malleable, sensitive to framing, and shaped by trust and justice concerns; they also highlight persistent gaps in geographic coverage and youth representation [17,27,28]. At the same time, the higher-education climate education literature shows both the promise and variability of instructional effects, suggesting the need for rigorous pre/post designs and theory-aligned outcome constructs [23,24]. Bringing these strands together implies a clear research gap: we still have limited empirical evidence on how structured university instruction that explicitly foregrounds regional inequality and governance dilemmas shapes students’ SRM cognition, perceived risks/benefits, and normative preferences—particularly in East Asian contexts where both climate vulnerability and geopolitical considerations may differ from Euro–American settings [28,29].

The present study at Zhejiang University contributes in three ways. First, it conceptualizes learning about solar radiation modification as multi-dimensional cognitive change, rather than a simple shift toward acceptance or rejection, thereby aligning measurement with how debates about legitimacy and responsible decision-making are commonly organized [34,35,45,46]. Additionally, it incorporates regional perspective-taking as an intentional pedagogical strategy to encourage reasoning that is sensitive to cross-regional differences in impacts and priorities, engaging with concerns that discussions of climate interventions can be shaped by a narrow set of experiences and assumptions [27,28,46]. Further, it also provides empirical evidence from an English-taught, university-wide climate change course at Zhejiang University, broadening the empirical base of SRM education and offering transferable implications for global oriented climate curriculum.

Based on the literature above, this study addresses three research questions: (RQ1) How does a structured classroom module affect students’ self-rated knowledge of SRM? (RQ2) How does instruction shift students’ support for SRM research versus support for SAI deployment and the framing of SRM as a “temporary bridge”? (RQ3) Which post-lecture beliefs (perceived benefits versus risks, confidence in governance, and moral-hazard concern) are associated with post-lecture support outcomes?

3. Materials and Methods

3.1. Course Context, Sampling, and Study Timeline

The study was embedded in an undergraduate elective course on global climate change offered at Zhejiang University. All enrolled students were invited to complete an anonymous survey immediately before the start of the SRM module (T1) and again after completion of the module (T2). The surveys were administered three weeks apart, reflecting the course cadence of one 120-min class per week and three consecutive sessions devoted to climate policy foundations and mitigation options, including SRM as a contingency response. Participation was voluntary and students could opt out at any time; surveys collected no directly identifying information. Across the three-week module, instruction combined short lectures with guided discussion and an in-class scenario exercise that required students to reason from a regionally situated perspective and to make explicit governance and equity assumptions. To provide an overview of the study design and intervention, Figure 1 summarizes the matched pre–post workflow and the key outcomes measured at each time point.

3.2. Educational Module Design and Implementation

The SRM educational module was designed to move beyond “technical facts only” instruction by explicitly foregrounding (i) regional heterogeneity in climate risks and SRM side effects, (ii) distributional inequality (who benefits, who bears risk, and how responsibility is allocated across regions and generations), and (iii) governance dilemmas (who decides, under what rules, with what safeguards and accountability). The module paired scientific content on mitigation pathways and SRM proposals with policy concepts such as fairness, common but differentiated responsibilities, precaution, and climate justice. Instructor-developed lecture slides (C5–C7) provided the core materials and guided students through a structured sequence of conceptual framing, evidence review, and applied deliberation.

Session 1 established the policy and governance framing for climate response options. It introduced climate governance as collective rule-making under the UNFCCC principles (e.g., fairness, common but differentiated responsibilities, and precaution) and linked vulnerability and resilience to structural inequalities in coupled human–environment systems. Students discussed why climate impacts and adaptive capacity differ across regions and why such asymmetries create persistent equity and legitimacy challenges for global climate action.

Session 2 focused on why mitigation lags despite increasing scientific urgency. Students engaged with exponential growth intuition, the planetary boundaries framework, and the logic of discounting in cost–benefit analysis. The lecture used the social cost of carbon to illustrate how normative choices (e.g., discount rates, whose damages count) shape policy conclusions and distributional judgments.

Session 3 reviewed mitigation policy instruments (command-and-control vs. market-based tools), co-benefits as a motivation and legitimacy strategy, and then introduced SRM as a contingency option when mitigation and adaptation alone may not avoid severe near-term risks. SRM content emphasized (a) the separation between temperature reduction and the root cause of climate change, (b) uncertainties and potential unintended consequences, (c) regionally uneven side effects (e.g., precipitation and monsoon disruption), and (d) governance questions such as “whose hand on the thermostat,” conflict risks, and moral hazard.

Required materials consisted of three instructor slide decks delivered across the three sessions (Session 1: Fundamentals of climate policy; Session 2: Exponential growth and the social cost of carbon; Session 3: Climate mitigation), which synthesized core concepts and empirical evidence in climate and economic policy, and introduced SRM as a contested policy option under uncertainty. To ensure that governance and distributional questions were explicitly foregrounded, the slides and in-class materials incorporated short excerpts and summaries from major SRM assessment and governance sources as well as peer-reviewed discussions of legitimacy, accountability, and responsible innovation. All materials were distributed to students via the course learning platform and served as the common evidence base for the structured exercise in Session 3.

The central activity was a structured perspective-taking exercise conducted in Session 3. Students were randomly assigned to small groups representing six world regions (East Asia, Southeast Asia, South America, North America, Europe, and Africa). Using an instructor-prepared regional profile handout and a common scenario prompt, each group selected a sub-region within its assigned region and worked through a shared set of questions: (1) What are the current and projected climate risks for your region under a baseline pathway (SSP2-4.5)? (2) Would you choose to implement stratospheric aerosol injection (SAI) as a response option? (3) If yes, what temperature target and start year would you propose, and what monitoring and exit criteria would you require? (4) If no, what risks or ethical/governance reasons motivate refusal? (5) How might your choice affect other regions, and what forms of consent, compensation, and accountability would be required to make the policy legitimate?

Groups produced a short decision memo summarizing their recommendation, the uncertainties they considered, and the governance conditions they regarded as necessary. A plenary discussion followed in which groups compared trade-offs across regions, explicitly highlighting distributional tensions (benefits vs. risks; present vs. future generations) and practical governance challenges of coordination, transparency, liability, and conflict prevention. The overall structure of the three-session SRM module, including the session topics, materials, in-class activities, and outputs, is summarized in

Table 1. SRM module schedule and instructional components (three sessions between T1 and T2).

Session	Topic	Reading/Materials	Activity (In-Class)	Output
S 1	Climate policy foundations; governance; vulnerability, resilience, and inequality	Slides C5: “Climate Mitigation and Adaptation & Fundamentals of Climate Policy” (includes UNFCCC principles; climate governance; vulnerability and structural inequalities; climate justice/equity)	Mini-lecture + guided discussion on governance principles (fairness, CBDR, precaution) and why vulnerability/adaptive capacity differ across regions	Participation notes; short written responses to “questions for thought” (optional)
S 2	Exponential growth; planetary boundaries; emission gaps; social cost of carbon and discounting	Slides C6: “Exponential Growth, Planetary Boundaries and Social Costs of Carbon” (includes SCC, discounting, damage functions; governance framing)	Concept questions + short calculations/discussion on discounting and how normative assumptions change policy conclusions	Completed worksheet/calculation notes; discussion summary
S 3	Mitigation policy instruments and co-benefits; SRM as contingency (Plan B); SRM risks, uncertainties, governance/ethics	Slides C7: “Climate Mitigation: Approaches & Policies” (includes mitigation instruments, co-benefits literature; SRM proposals; regional impacts; governance/ethics; reflective.org simulator; SRM scenario game prompts)	Structured perspective-taking “SRM Scenario Game” in regional groups + plenary comparison of trade-offs and governance conditions	Group decision memo + plenary reflection; administered post-survey (T2) after session

.

3.3. Participants, Survey Administration, and Ethics

All 106 students enrolled in the course were invited to complete two anonymous surveys: a pre-lecture survey administered immediately before the SRM module and a post-lecture survey administered immediately after the module. Participation was voluntary, and students were informed that (i) the survey was for research/educational evaluation purposes, (ii) they could stop at any time, and (iii) responses would be analyzed only in aggregated, de-identified form. To enable within-person analyses without collecting personal identifiers, students created a self-generated anonymized participant number at T1 and re-entered the same number at T2; the code was used solely for matching pre/post responses and was not linked to any identifying information.

Of the 106 enrolled students, 103 provided valid matched responses after applying the exclusion rules described below and constitute the analytical sample (N = 103).

3.4. Measures and Data Processing

Both surveys used ordinal response scales. Key outcomes measured pre and post were: (1) self-rated SRM knowledge (1–3); (2) support for SRM research (1–5); (3) support for SAI deployment (1–5); (4) agreement with framing SRM as a “temporary bridge” (1–5); and (5) policy priority ranging from low-carbon development (1) to climate intervention (5). Selected pre-survey items included a “Don’t know” option. The pre survey also captured major field of study and baseline climate-policy attitudes. The full pre- and post-module questionnaires are provided in Appendix A.

We applied pre-specified response-quality rules prior to outcome analyses and removed excluded cases from the analytical dataset. Responses were excluded if they (i) could not be matched across pre and post surveys using the anonymized participant number; or (ii) exhibited straightlining across the Likert-scale items (operationalized as selecting the exact same response option across all 1–5 items within a survey, with no variation) and provided no substantive explanation in any open-ended response. In total, three cases were removed, yielding N = 103 matched valid participants for analysis. These exclusion criteria were specified before inspecting pre–post outcome differences; the study was not preregistered.

For items where “Don’t know” was available (coded as 6 in the dataset), these responses were treated as missing for numeric summaries and inferential tests, but their frequencies are reported descriptively. For each pre–post test, matched pairs were included only when both the pre and post response were non-missing for that outcome.

3.5. Quantitative Analysis

We report means with standard deviations, medians with interquartile ranges, and “Don’t know” response rates for ordinal items to summarize central tendency and uncertainty. Within-person pre–post differences were tested using two-sided Wilcoxon signed-rank tests. To control the family-wise error rate across the five key outcomes, p-values were adjusted using the Holm procedure; we report both raw and Holm-adjusted p-values for transparency. For effect size reporting, we present Cohen’s dz (paired standardized mean change) as an approximate standardized change indicator that is commonly reported in education research for interpretability, while also reporting a complementary nonparametric effect size r based on the Wilcoxon z-statistic.

To examine correlates of post-lecture support, we estimated ordered logistic regression models (proportional-odds) given the ordinal nature of the dependent variables. These models are interpreted as descriptive associations rather than causal effects. As robustness checks for model specification, we also estimated multinomial logit models that relax the proportional-odds constraint; results were substantively consistent with the ordered-logit estimates in sensitivity analyses. We assessed the proportional-odds assumption using standard diagnostics. Diagnostics indicated no substantive violations, and multinomial-logit specifications yielded the same substantive conclusions for the key predictors; we therefore report the ordered-logit results for parsimony.

3.6. Qualitative Coding of Open-Ended Responses

Post-lecture open-ended responses were collected from Question 10, which asked students to state the main reason for any change in their view, or—if their view did not change—the single most important condition shaping their support for or opposition to SRM. Responses could be written in either Chinese or English.

We used a mixed deductive-inductive thematic approach. We first defined six higher-order themes a priori, based on recurrent dimensions in SRM perception and governance debates and on constructs emphasized in the course module: (1) feasibility/cost/technical readiness; (2) risks and unintended consequences; (3) moral hazard and mitigation priority; (4) governance and accountability; (5) temporary or emergency use; and (6) ethics/fairness/distribution. We then conducted a pilot pass through a subset of responses to clarify inclusion/exclusion rules and to expand a bilingual (Chinese-English) keyword list capturing common synonyms and expressions. The final keyword dictionary was applied to the full set of responses, and each response could receive multiple theme tags because students often referenced more than one consideration in a single answer.

Because dictionary-based tagging captures the salience of explicit mentions rather than latent meanings or argument structure, we interpret theme frequencies as descriptive indicators of what students most often foregrounded in their short answers, and we use these patterns to contextualize the quantitative results rather than as stand-alone qualitative inference.

4. Results

4.1. Sample Characteristics

Table 2. Major field of study (pre survey; N = 103).

Major	N	%
Natural Sciences	21	20.4
Engineering	16	15.5
Business and Economics	2	1.9
Medicine and Health Sciences	7	6.8
Arts	2	1.9
Computer Science	1	1.0
Environmental Sciences	24	23.3
Law	6	5.8
Agriculture and Forestry	11	10.7
Other	13	12.6

Notes: Percentages are of N = 103.

summarizes respondents’ self-reported major field of study from the pre survey. The cohort was academically diverse, but it skewed toward STEM fields: Environmental Sciences was the single largest category (23.3%), followed by Natural Sciences (20.4%) and Engineering (15.5%), with STEM-related majors comprising at least ~78% of the sample. This disciplinary composition provides important context for interpreting baseline knowledge and attitudes and for generalizability to non-STEM cohorts.

4.2. Baseline Climate and Policy Attitudes

At baseline, most respondents viewed climate change as a very or extremely serious problem (72.8% selected 4–5 on a 1–5 seriousness scale). Support for carbon pricing was also relatively high (62.1% selected 4–5 on a 1–5 carbon tax support scale). In contrast, most students disagreed that market forces alone can decarbonize the global economy by 2050 (76.7% selected 1–2 on a 1–5 agreement scale). These baseline attitudes indicate a cohort that is generally climate-concerned and policy-engaged, providing a relevant context for interpreting SRM judgments.

4.3. Pre–Post Changes in SRM Knowledge and Attitudes

Figure 2 visualizes pre and post response distributions for key outcomes. Overall, the module substantially increased self-rated knowledge, while shifting support away from deployment-oriented positions.

Table 3. Pre vs. post descriptive statistics for key outcomes.

Outcome	Pre Mean (SD)	Pre Median [IQR]	Pre DK%	Post Mean (SD)	Post Median [IQR]	Post DK%
Self-rated knowledge (1–3)	1.86 (0.51)	2 [0]	—	2.33 (0.53)	2 [1]	—
Support for SRM research (1–5)	3.76 (0.88)	4 [1]	1.9	3.54 (1.05)	4 [1]	0.0
Support for SAI deployment (1–5)	3.42 (0.90)	3 [1]	5.8	2.95 (0.98)	3 [2]	0.0
Support: SRM as temporary bridge (1–5)	3.43 (1.06)	4 [1]	1.0	3.12 (1.14)	3 [2]	0.0
Priority: Low-carbon (1) to intervention (5)	3.08 (1.03)	3 [2]	2.9	2.40 (1.09)	2 [1]	0.0

Notes: Self-rated SRM knowledge is measured on a 1–3 scale; all other outcomes use 1–5 Likert scales (higher = more support/agreement). Means/medians exclude “Don’t know”. DK% = percent selecting “Don’t know”. — = not applicable because the self-rated knowledge item did not include a “Don’t know” option.

reports descriptive statistics, and

Table 4. Within-person change and effect size.

Outcome	N Pairs	Mean Δ	p (Raw)	p (Holm)	Cohen’s dz	r (Wilcoxon)
Self-rated SRM knowledge (1–3)	103	0.47	1.93 × 10−8	9.64 × 10−8	0.67	0.55
Support for SRM research (1–5)	101	−0.24	0.1191	0.1191	−0.16	−0.16
Support for SAI deployment (1–5)	97	−0.46	0.0028	0.0084	−0.33	−0.27
Agreement: SRM as a temporary bridge (1–5)	102	−0.30	0.0585	0.1169	−0.19	−0.17
Policy priority: low-carbon (1) → intervention (5)	100	−0.68	2.50 × 10−5	0.0001	−0.45	−0.40

Notes: Mean Δ = Post − Pre. Positive values indicate higher post-lecture responses. Cohen’s dz is reported as an approximate standardized paired change.

reports within-person change tests (Wilcoxon signed-rank) with both raw and Holm-adjusted p-values and effect sizes.

Consistent with the descriptive shift in Figure 2, students entered the module with moderate openness to the idea of deployment (pre mean support for SAI deployment = 3.42) but became more cautious after learning about uncertainties, risks, and governance constraints (post mean = 2.95). The mean decline of −0.46 on a 1–5 scale corresponds to nearly half a response category, implying a practically meaningful shift from “somewhat support” toward a more neutral/hesitant stance at the cohort level. At the same time, support for SRM research remained relatively high (post median = 4), indicating a common distinction between valuing research as information-gathering versus endorsing real-world experimentation or deployment. To further quantify these shifts, Figure 3 reports mean post–pre differences with 95% confidence intervals.

4.4. Post-Lecture Beliefs: Literacy, Benefits vs. Risks, Governance, and Moral Hazard

In the post survey, SRM conceptual literacy was high: 86.4% selected the correct SRM definition in a multiple-choice item. Perceptions of benefits versus risks were centered near “about equal” (median = 3). Confidence in international governance capacity was moderate (median = 3), while concern about moral hazard leaned toward agreement (median = 4 among non-missing responses). Figure 4 shows the post-lecture distributions for these items.

4.5. Correlates of Post-Lecture Support

We examined correlates of post-lecture support using ordered logistic regression, which is appropriate for the ordinal 1–5 outcomes.

Table 5. Multivariable correlates of post-lecture support (ordered logit; odds ratios, 95% CI).

Predictor (Standardized)	Post Support: Research	Post Support: SAI Deployment	Post Support: Temporary Bridge
Baseline support	0.76 [0.50, 1.14]	0.79 [0.52, 1.20]	0.72 [0.50, 1.06]
Baseline climate concern	0.93 [0.62, 1.40]	0.90 [0.60, 1.36]	1.55 * [1.05, 2.30]
Post self-rated SRM knowledge	1.21 [0.81, 1.81]	0.70 [0.45, 1.11]	1.15 [0.76, 1.72]
Perceived benefits vs. risks (post)	1.78 * [1.11, 2.85]	2.16 ** [1.35, 3.47]	2.57 *** [1.60, 4.11]
Governance confidence (post)	1.72 * [1.08, 2.74]	1.44 [0.89, 2.34]	0.66 [0.43, 1.01]
Moral-hazard concern (post)	1.57 * [1.06, 2.34]	1.09 [0.71, 1.67]	1.56 * [1.05, 2.33]
Correct SRM definition	3.23 [0.94, 11.04]	1.10 [0.34, 3.55]	1.00 [0.34, 2.99]

Notes: Entries are odds ratios (OR) per 1 SD increase in the predictor (binary for SRM definition), with 95% confidence intervals in brackets. Stars: * p < 0.05, ** p < 0.01, *** p < 0.001 (two-sided). Ordered logistic regression estimated with proportional-odds specification; results are descriptive associations. N = 97 (research), N = 93 (SAI), N = 98 (temporary bridge).

reports odds ratios (OR) with 95% confidence intervals. Across all three outcomes, perceived benefits versus risks was the most consistent correlate of support: higher perceived net benefits were associated with higher odds of stronger support for SRM research (OR = 1.78), SAI development and deployment (OR = 2.16), and the temporary-bridge framing (OR = 2.57). Governance confidence was positively associated with research support (OR = 1.72) and marginally negative for the temporary-bridge outcome (OR = 0.66). Moral-hazard concern was positively associated with stronger support for SRM research (OR = 1.57) and for the temporary-bridge framing (OR = 1.56), indicating that some respondents who endorsed research or conditional use simultaneously attended to moral-hazard arguments. Baseline climate concern was associated with greater support for SRM as a temporary bridge (OR = 1.55). Model N varies modestly across outcomes due to “Don’t know” responses on baseline or moral-hazard items. As shown in Figure 5, perceived benefits versus risks was the most consistent correlate of stronger post-lecture support across outcomes.

4.6. Themes in Open-Ended Responses

Table 6. Theme mentions in post-survey open responses (keyword-based coding; N = 103).

Theme	N (of 103)	%
Feasibility/cost/technical readiness	39	37.9
Risks and unintended consequences	30	29.1
Moral hazard and mitigation priority	23	22.3
Governance and accountability	22	21.4
Temporary/emergency use	13	12.6
Ethics/fairness/distribution	4	3.9

Notes: Responses can mention multiple themes; percentages therefore do not sum to 100%.

summarizes theme mentions in post-survey open-ended responses. The most common themes involved feasibility/cost/technical readiness and concerns about risks and unintended consequences. Governance and moral hazard were also frequently referenced, consistent with the quantitative emphasis on perceived benefits versus risks and governance confidence.

Illustrative anonymized excerpt included: (i) “SRM’s risks are not yet clear; it may affect ecosystems.” (risks/unintended consequences); (ii) “SRM cannot replace emissions cuts; it is a stopgap measure.” (moral hazard/mitigation priority); and (iii) “The key condition is a strong global governance framework to ensure fair and safe deployment.” (governance/accountability).

5. Discussion

5.1. From Awareness to Caution: Instruction Reshapes SRM Judgments

The most robust change observed was an increase in self-rated SRM knowledge after the classroom module (dz = 0.67). At the same time, attitudes shifted toward greater caution: support for SAI deployment declined and policy prioritization moved toward low-carbon development rather than climate intervention. These patterns mirror prior survey and deliberative findings that baseline SRM familiarity is typically low and that judgments are contingent and sensitive to governance, perceived risks, and scenario specificity; additional information and deliberation often increase conditionality or caution rather than producing a linear “knowledge to acceptance” trajectory [16,17,18,29,45,46]. In this sense, the module appears to have moved students from abstract openness toward a more governance-aware evaluation under uncertainty.

Substantively, many students appeared to begin the module with an “engineering optimism” intuition—SRM might be deployable as a response to escalating climate risks. After engaging with the module’s content on physical uncertainties, ecological side effects, and institutional/governance constraints—especially through the regional scenario exercise that required students to articulate assumptions about uneven impacts and decision-making—students more frequently emphasized unintended consequences and system-level risks, contributing to reduced deployment support. This direction of change is consistent with deliberative studies in which support becomes more conditional (or declines) once participants confront governance feasibility and distributional trade-offs in concrete terms [45,46].

5.2. “Research Yes, Deployment No”: Distinguishing Epistemic Value from Risk Acceptance

A key insight is the divergence between relatively stable, moderately positive support for SRM research and declining support for SAI deployment. Even after instruction, students’ median support for SRM research remained at 4 (on a 1–5 scale), suggesting that most respondents viewed research as valuable for improving understanding and informing governance. However, the decline in deployment support indicates reluctance to accept the real-world risks and potential irreversibility associated with large scale intervention. This pattern aligns with prior SRM perceptions and governance scholarship, which often finds higher openness to research than to deployment and emphasizes conditional support based on governance safeguards, transparency, and perceived necessity [4,7,16,17,45].

This “research–deployment gap” is important for policy and communication: educational interventions may increase literacy without producing blanket enthusiasm. Instead, instruction can clarify that supporting research is compatible with opposing, constraining, or deferring deployment—especially under uncertain governance conditions. Our post-lecture regressions are consistent with this interpretation: perceived benefits versus risks was the strongest correlate of support for deployment and the temporary-bridge framing, highlighting the centrality of risk–benefit judgments in the classroom context—a pattern also reported in broader perceptions research [17,29,46]. More generally, calls for social-science research on solar geoengineering emphasize the value of empirically identifying these pathways (risk perceptions, governance confidence, moral-hazard concerns) rather than assuming a simple deficit model [48].

5.3. Preference for Low-Carbon Development and the Salience of Moral Hazard

The significant shift of the priority item toward low-carbon development suggests that, when presented with SRM as a policy option, students generally preferred mitigation-focused pathways. This aligns with the moral-hazard/mitigation-deterrence concerns emphasized in SRM governance debates and with open-ended responses that explicitly framed SRM as unable to substitute for emissions reductions [41,43]. Together, these results support the interpretation that students perceived SRM as, at most, a constrained and potentially temporary tool, rather than a desirable strategy to pursue proactively—echoing recent governance debates that emphasize strict constraints and strong safeguards around deployment [12].

Notably, moral-hazard concern was positively associated with support for SRM research. One plausible interpretation is that students who support research may also be more attentive to the broader governance discourse—recognizing moral hazard as a reason to design research governance carefully rather than a reason to avoid knowledge production entirely. This possibility underscores the value of measuring multiple, conceptually distinct outcomes (knowledge, risk–benefit reasoning, governance confidence, and moral-hazard concern) when evaluating SRM education [48]. Future work could disentangle whether this pattern reflects “informed caution” versus other motivational pathways.

5.4. Implications for SRM Education and Deliberative Course Design

Taken together, the pattern of results suggests that SRM education is not well captured as a simple “more information to more support” trajectory. The module increased self-rated knowledge while support for SAI deployment and political prioritization of climate intervention declined, and students’ open-ended responses frequently foregrounded feasibility constraints, unintended consequences, governance conditions, and moral-hazard concerns. Mechanistically, the shift plausibly reflects specific module elements: (i) uncertainty and risk-focused scientific content can move students from abstract “engineering optimism” to a more precautionary stance; (ii) the regional role-play/scenario exercise makes heterogeneous impacts and equity trade-offs salient; and (iii) governance framing highlights the real-world constraints of decision-making under fragmented international institutions and contested legitimacy [7,36,37,54]. This combination is consistent with SSI-oriented teaching approaches that treat learning as integrating evidence evaluation with moral and political reasoning under uncertainty [51,52,53].

Accordingly, SRM curricula may be most informative when they integrate technical content with structured opportunities to reason about uncertainty, distributional trade-offs, and institutional constraints, rather than presenting SRM as a purely technical option. Notably, although the module explicitly foregrounded distributional inequality, explicit fairness or distributive-justice language appeared relatively infrequently in students’ short open-ended answers. This suggests a concrete design improvement: future SRM teaching can add more dedicated prompts and worked examples (e.g., consent, compensation, intergenerational equity, and “who pays/who benefits” scenarios) to help students translate governance-aware evaluation into clear distributive-justice reasoning. For educators, a practical implication is to treat teaching about SRM as an opportunity to help students separate (i) epistemic questions (what do we know and what is uncertain?), (ii) normative questions (who bears risks and who benefits?), and (iii) institutional questions (who decides and under what safeguards?). Such separation may help students articulate the common stance observed here: SRM research can be worthwhile, but the risks of deployment are not something most students want to accept, especially when low-carbon development remains a preferred and more controllable strategy.

Finally, we note that the instructional design also aligns with relational pedagogies that place instructor–student and peer relationships at the center of learning. The module’s small-group regional role assignment and facilitated plenary discussion were intended to support dialogic co-construction of understanding, responsiveness to learners’ identities and contexts, and explicit attention to power and justice in SRM governance debates.

5.5. Limitations and Future Research

The findings should be interpreted in light of the study’s context and scope. The evidence comes from a single classroom cohort (N = 103 matched) without a randomized control group, so the results are best read as descriptive within-setting change rather than causal effects that can be generalized broadly. The study was conducted in one elective, English-taught course at Zhejiang University. Students self-selected into the elective course, which may introduce selection effects (e.g., higher baseline interest or climate concern) and further limit generalizability. The sample skews toward STEM majors (notably Environmental Sciences). This setting is substantively meaningful, but it also suggests that baseline knowledge, climate concern, and receptivity to SRM content may differ in non-STEM cohorts, other universities, non-English-taught courses, or different cultural and political environments. Several key constructs were measured with single-item Likert questions, and SRM knowledge was self-rated; future work would be strengthened by validated multi-item scales and objective knowledge measures (e.g., true/false/”don’t know” items) to triangulate perceived learning. Finally, outcomes were assessed immediately after the module; longer-term follow-ups are needed to test whether knowledge gains and attitude shifts persist and how they evolve with continued exposure to climate policy debates.

Future research could extend this design by incorporating comparison groups, collecting delayed post-tests, and replicating the module across multiple institutions and more diverse disciplinary mixes. Qualitative follow-ups (e.g., interviews or deliberative focus groups) could also clarify why students often separate support for research from reluctance toward deployment and how governance concerns and moral-hazard reasoning develop over time.

6. Conclusions

Using matched pre–post surveys in a university climate change course (106 enrolled; 103 valid matched responses), we find that SRM education can increase perceived knowledge while simultaneously sharpening reservations about deployment and political prioritization. Figure 6 provides a visual synthesis of the study design, the structured SRM module, and the observed pre–post shifts in key outcomes alongside the main post-module themes. Three conclusions stand out:

(1)

Self-rated SRM knowledge increased substantially after the classroom module (dz = 0.67).

(2)

Support for SAI deployment declined (dz = −0.33), and students shifted toward prioritizing low-carbon development over climate intervention (dz = −0.45).

(3)

Most students continued to view SRM research as valuable, but expressed reluctance to accept SRM-related risks and emphasized governance, moral hazard, and unintended consequences.

Taken together, the results emphasize that growing SRM literacy does not necessarily translate into deployment enthusiasm. Instead, education may cultivate a “cautious engagement” stance—supporting research and governance exploration while preferring mitigation-centered pathways as the primary response to climate change.

More broadly, beyond documenting within-classroom changes in SRM perceptions, this study contributes to sustainability by identifying how structured climate-intervention instruction can strengthen students’ capacity to weigh trade-offs among mitigation, governance, and technological risk under uncertainty. By foregrounding regional heterogeneity, legitimacy, and moral-hazard concerns, the module supports sustainability-oriented decision-making competencies that are relevant for anticipatory governance of emerging climate interventions and for education aimed at sustainable development. The observed shift toward mitigation-centered prioritization alongside continued support for research highlights a form of “cautious engagement” that aligns with sustainability goals emphasizing risk reduction, equity, and accountable governance.