Scientists as communicators: A randomized experiment to assess public reactions to scientists’ social media communication along the science-advocacy continuum INTRODUCTION The queson of what roles, if any, sciensts should play in the dialogue about public policy is a perennial point of discussion in the scienfic community. It is oſten assumed that policy advo- cacy – showing support for one policy approach over another – is likely to compromise the per - ceived credibility of the individual scienst engaged in such behavior, and may also harm the credibility of the scienfic community more broadly (Nelson & Vucech, 2009). Donner (2014) argues that a binary view of advocacy – where statements are either objecve or advocacy – is too simplisc, and proposes that public communicaon by sciensts should instead be thought of as falling along a connuum in terms of its “level of advocacy” based upon the extent of nor - mave judgment implicit in a statement. He predicts that as the extent of normave judgment increases, it poses a relavely greater risk to a scienst’s perceived credibility. Research suggests that sciensts are judged as more credible and trustworthy to the extent that they are seen as being unbiased and having posive intenons (Eagly, Chaiken, & Wood, 1981; Fiske & Dupree, 2014). Relevant to the debate about advocacy, we examine percepons that the scienst’s statements are aributable to their evaluaon of scienfic evidence and/or their polit - ical views. We also examine the extent to which their statement is perceived as intended to per - suade the public to take acon or intended to provide imparal informaon. Such aribuons may be associated with differences in the perceived credibility of sciensts that engage in differ - ent levels of advocacy or with differences in the percepon of the trustworthiness of the broad- er community of sciensts as well as public support for funding research. Finally we also examine the role of polical ideology in shaping audience responses to scien- sts’ statements. In the United States, atudes about climate change are polarized along par- san and ideological lines with Liberals tending to be more accepng of the existence of climate change and supporve of societal acon compared to Conservaves, and polical idenficaon has been found to condion informaon processing of messages about climate change (Gromet, Kunreuther, & Larrick, 2013; Hart & Nisbet, 2012; McCright & Dunlap, 2011). RESULTS A three-way MANOVA examining the influence of level of advocacy, type of scienst, and poli- cal ideology (liberal, moderate, or conservave) on our DVs was conducted to answer our re- search quesons. Mulvariate tests found a significant main effect of level of advocacy F(20, 4748)=10.38, p<.001, Pillai’s Trace=.168, paral η2 = .042, and a significant main effect of ideology F(10, 2370)=4.90, p<.001, Pillai’s Trace=.041, paral η2 = .020. There was no significant main effect of type of scienst F(5, 1184)=1.12, p=.346, Pillai’s Trace=.005, paral η2 = .005. There were no significant two-way or three-way interacons among the independent variables. Post-hoc main comparisons for each of the DVs were conducted using Tukey’s HSD to correct for family-wise er - ror rates. RQ1: Will statements containing progressively higher levels of advocacy be associated with dif- ferent evaluaons of the scienst in terms of the extent to which their statement is: a) more aributed to their polical views The control condion is less aributed to the scienst’s polical views than all other condions ex - cept for Risks & Impacts (p<.01, see Figure 1). The Risks & Impacts statement is also less aributed to polics than the Specific Acon statement (p<.05). b) less aributed to their evaluaon of scienfic evidence The Specific Acon statement is less aributed to science than all other condions except for Policy Opons & Consequences (p<.01, see Figure 1). We also found that the control condion is more highly aributed to science than the Policy Opons & Consequences statement (p<.01). c) viewed as more intended to persuade the public to take acon to address climate change The Non-specific and Specific Acon statements were seen as more intended to persuade relave to all other levels of advocacy (p<.01, see Figure 2). d) less intended to provide imparal informaon about climate change The Non-specific and Specific acon statements were seen as less intended to inform relave to the control and Risks & Benefits condions (p<.05, see Figure 2). Furthermore, the control condi- on was seen as more intended to inform than all other levels of advocacy (p<.05). DISCUSSION The results presented here appear to challenge widely held assumpons about some of the risks associated with sciensts engaging in policy advocacy. Contrary to convenonal wisdom, our study suggests there is not a linear associaon between a scienst’s level of advocacy and their perceived credibility. Although we found that when a scienst endorsed a specific policy to address climate change their credibility was relavely lower, in absolute terms, their perceived credibility was sll above the midpoint on the scale. Furthermore, we found that ad- vocang for a non-specific acon was not associated with a decrease in credibility. There also does not appear to be a simple trade-off between persuasive and informave in- tenons, and polical and scienfic aribuons. Across condions, most levels of advocacy were seen as concurrently intended to persuade and inform, and aributable to both scien- fic and polical movaons. For the issue of climate change, it appears that statements ad- vocang acon without endorsing specific policies are seen as parcularly movated by science and intended to persuade the public to take acon. Although Conservaves evaluated the sci- enst more crically than Moderates and Liberals, this difference was independent of the level of advocacy. Importantly, we did not observe negave spillover effects of the individual scienst’s advo- cacy on atudes toward the broader scienfic enterprise. Both the perceived trustworthiness of the scienfic community they were intended to represent as well as support for funding cli- mate change research were unaffected by the level of advocacy. While the results presented here may give some sciensts more confidence that their public statements about climate poli- cy carry less professional risk than previously thought, we urge cauon in over interpreng the external validity of our results. More research is needed to examine policy advocacy in addi- onal communicaon contexts beyond single, brief posts on social media to include other for - mats such as newspaper arcles, op-eds, and video interviews. e) result in lower perceived credibility of the scienst? The Specific Acon statement was viewed as less credible than all other condions except Poli- cy Opons & Consequences (p<.05, see Figure 3). RQ2: Will there be a spillover effect such that high levels of advocacy will be associated with a) lower levels of general trust toward climate sciensts & television weathercasters more broadly, and b) lower levels of support for funding research to study climate change? No significant main effects of level of advocacy were observed on general trust toward climate sciensts, F(4, 602)=.410, p=.802; nor on general trust toward television weathercasters F(4, 581)=.247, p=.969. Lastly, there was no main effect of level of advocacy on support for funding research to study climate change, F(4, 1185)=1.93, p=.103. RQ3: Will the effects of level of advocacy in RQ1 and RQ2 be condioned by the respondent’s polical ideology? Polical ideology was coded and modeled as three categories: Liberal, Moderate, and Conser - vave. There were no significant differences between Liberals and Moderates across all of the DVs. However, Conservaves were less likely to aribute the expert’s statement to an evalua- on of the science compared to both Liberals and Moderates, regardless of level of advocacy (p<.001). Conservaves were also more likely to aribute the statement to the scienst’s poli- cal views than Liberals and Moderates (p<.05). Conservaves were less likely to see the state- ments as intended to inform compared to Liberals and Moderates (p<.01). However, there were no significant differences across polical ideology with regard to the percepon that the statements were intended to persuade people to take acon. Finally, Conservaves also viewed the scienst as less credible relave to Liberals and Moderates (p<.001). There were no interac- ons between level of advocacy and ideology. John Kotcher ([email protected]), Emily Vraga, Teresa Myers, Neil Stenhouse, Connie Roser-Renouf, and Edward Maibach Department of Communicaon, George Mason University METHOD Parcipants • Online survey experiment in Qualtrics • Data collected July 30-August 5, 2014 • Quota Sample from Survey Sampling Internaonal to match U.S. demographics • (n=1235) ~100 parcipants per condion Procedure • 5 (level of advocacy) x 2 (type of scienst) factorial design • Randomly assigned to read one ficonal Facebook post then answered quesons Levels of Advocacy 1. Control • Menons 400ppm CO 2 milestone, no discussion of risks 2. Risks & Impacts • Describes public health risks of climate change 3. Policy Opons & Consequences • Discusses pros & cons of two migaon policies (see specific acon below) 4. Non-specific Acon • Urges acon to address climate change w/o specific policy menoned 5. Specific Acon (2 condions collapsed for analysis) • Liberal congruent policy – Endorses limits on CO 2 at coal plants • Conservave congruent policy –Endorses more nuclear plants Dependent Measures Aribuon to polical views • Dr. Wilson’s statement was shaped by his polical views about climate change. • (1=Strongly disagree, 7=Strongly agree.) Aribuon to scienfic evidence • Dr. Wilson’s statement was shaped by his evaluaon of the scienfic evidence about cli- mate change. • (1=Strongly disagree, 7=Strongly agree.) Perceived goal to persuade • The goal of the Facebook post was to persuade people to take acon to address climate change. • (1=Strongly disagree, 7=Strongly agree.) Perceived goal to inform • The goal of the Facebook post was to provide imparal informaon about climate change • (1=Strongly disagree, 7=Strongly agree.) Perceived credibility • Adapted from McCroskey & Teven (1999); • Mean of 9 semanc differenal items (α=.91; Min=1, Max=7) • Example: Not at all trustworthy—Extremely trustworthy