Knowledge and attitudes about deforestation: testing strategies to change attitudes
Ross Taplin
A *
A
Abstract
With increasing concerns about global climate change and deforestation, attitudes towards environmental protection are important, especially in democracies where citizens can influence policy direction.
This paper investigates people’s knowledge of a local environmental issue, deforestation of an ecologically diverse forest. Strategies to change environmental attitudes are also tested.
An online survey measured people’s generic knowledge and technical knowledge regarding deforestation due to bauxite mining. It also contains a randomised experiment (24 factorial design) to test whether four pieces of information cause changes in attitudes towards deforestation due to mining. The method incorporates the practical reality that environmental protection is often in conflict with other objectives such as economic prosperity, so investigates the issue more objectively than only investigating changes in one direction on a spectrum of protection or development.
Respondents have high generic knowledge but low technical knowledge, suggesting the need for further education. Information can change attitudes for or against environmental protection, and narratives are more successful than images. However, results vary depending on how attitudes are measured.
The paper demonstrates how identifying misconceptions or knowledge gaps is an important step towards changing attitudes about the need for environmental protection.
Implications include evidence of which strategies can influence public attitudes towards, and away from, environmental protection.
Keywords: attitudes, bauxite, climate change, deforestation, ecology, environment, jarrah, knowledge, randomised experiment, restoration.
Introduction
Anthropogenic climate change is predicted to at best result in climate extremes with catastrophic cost to the environment and humans (Trisos et al. 2020). Nevertheless, many people still hold attitudes that climate change is not real or an important issue (La et al. 2024), especially compared to more immediate concerns such as jobs, cost of living and economic prosperity. This was evident in the recent presidential election in the USA where Donald Trump, now President, was democratically elected after explicitly promoting policies that are detrimental regarding climate change.
Attempts by scientists and others to find remedies to combat climate change (such as reducing emissions by replacing fossil fuel consumption with renewables) are severely hampered when the public, and their elected governments, promote alternative agendas. It is unlikely people actively desire environmental destruction just for the sake of it, but they value other objectives such as economic prosperity in the form of jobs and revenue more highly in a consumerism society where the environment can be seen as another resource to be consumed for financial benefit.
Hence, public attitudes to environmental protection matter. Research documenting these attitudes is therefore important. Arguably more important is research that aims to change these attitudes; for example, by providing information to correct misconceptions and increase awareness and knowledge. Mendy et al (2024) provide a systematic literature review of climate change deniers and how interventions including education and message framing might counteract denialism. They conclude further research is required; while research on changing attitudes is growing, it still lacks empirical studies on the effectiveness of interventions. This paper examines these related research questions:
RQ1: How knowledgeable are people about environmental issues, both generic facts (such as whether the label ‘diversity’ applies to a specific environment) and technical (such as whether they can distinguish between diverse and non-diverse forests)?
RQ2: What are people’s attitudes to environmental protection and what information changes these attitudes?
Research question RQ2 considers not only correlations between attitudes and demographics, but causal effects from a randomised experiment. This is crucial when developing strategies to change attitudes. The study focusses on deforestation of a diverse forest close to a major population centre. Results suggest people are mostly aware of generic facts but unaware of technical details, even when the study site is well known and close to where they live. Some information provided within the randomised experiment causes changes in attitudes, but these effects vary depending on how the attitude is measured and the person’s demographics.
Context and hypotheses
Literature on attitudes to environmental protection is vast and spread over multiple disciplines (Mendy et al 2024). Within science, the recent literature review of some 450 pieces of research on ecosystem adaption by Vella et al. (2021) identified five challenges: (1) scientific conflicts and debates over the ‘facts’; (2) social challenges; (3) governance challenges; (4) epistemic challenges; and (5) ontological conflicts. This recognises the importance of not only attitudes in society, but also the role of facts and knowledge that form these attitudes.
The scientific literature often examines restoration of damaged environments, which is important when damage has already occurred or expected due to events such as climate change. However, borrowing from medicine the catchphrase ‘prevention is better than cure’, limiting environmental damage in the first place is also important. One example of prevention rather than cure is limiting deforestation. As summarised in a review of the biophysical mechanisms by which forests influence climate, Lawrence et al. (2022, p. 2) state: ‘Protection, expansion, and improved management of the world’s forests represent some of the most promising natural solutions to the problem of keeping global warming below 1.5–2 degrees (Griscom et al. 2017; Roe et al. 2019).’ Achieving this is complicated because deforestation brings economic benefits including revenue from timber, agriculture and mining. When attempts are made to regenerate forests (for example after mining), the extent to which this is successful is questionable (Campbell et al. 2024).
The scientific literature that focuses on people’s attitudes (rather than biology or environmental science of ecosystems) tends to focus on attitudes to climate change and environmental protection. The World Heritage Listed Great Barrier Reef (GBR) is an example that has attracted research on public attitudes. Lockie et al. (2024) and Curnock et al. (2019) compare attitudes to environmental issues for the GBR at two points in time. Curnock et al. (2019) attribute changes in attitudes to the Great Barrier Reef between 2013 and 2017 to mass coral bleaching, and while this is very plausible other changes between these time points might provide alternative explanations. Hence, the scientific superiority of randomised experiments, when practical, to confidently make conclusions of causality regarding why attitudes change. Other international examples include Zeng et al. (2023) who similarly compared attitudes at different times in China quantitatively and Boakes et al. (2023) who used qualitative interviews to investigate how coral reef restoration programs can change environmental attitudes in Indonesia.
The literature also contains research investigating how to change attitudes and behaviours. Steg and Vlek (2009) review the literature in psychology and suggest a future research agenda, arguing psychology has a place in promoting pro-environmental behaviours. Scholars from marketing, science and other areas can also contribute. For example, to examine the effect of how information is presented Maynard et al. (2024) randomised respondents to see either a social marketing poster or a scientific poster and found the social marketing poster was superior at promoting pro-environmental attitudes. Fesenfeld et al. (2024) provides a recent literature review of 121 experimental studies, primarily relating to issues of climate change, with a focus on interpreting how different people may be impacted by treatments such as messages.
Research using randomised experiments to make causal conclusions have investigated treatments such as relabelling climate change as climate crisis (Hung and Bayrak 2020), providing general information rather than specific case studies using thematic versus episodic frames (Hart 2011), and highlighting risks of failing to combat climate change with benefits of climate mitigation (Bain et al. 2012, 2016; Bernauer and McGrath 2016). Bolsen et al. (2019) provide evidence the use of visual imagery such as maps can change attitudes to climate change, with a particular focus on reducing political polarisation (especially Democrat vs Republican) of attitudes on climate change within the USA.
Literature specifically on attitudes towards deforestation is sparse. In the context of forest conservation in Brazil, Bakaki and Bernauer (2016) found that locals have attitudes that do not support protection of forests (indeed, their abstract instead asks ‘What could other (richer) countries do, in this context, to encourage forest conservation in Brazil and other tropical developing countries?.’ (Bakaki and Bernauer (2016), p. 1). This rather pessimistic conclusion was based on local economic pressures; locals need the economic benefits from agriculture, which requires the deforestation. Erikson and Klapwijk (2019) examine attitudes to deforestation in Sweden, such as preferences between carbon substitution and biodiversity conservation, including relationships between demographics such as respondent’s knowledge and attitudes. Neither utilise randomised experiments to test causal effects.
Based on the above literature, research question RQ2 not only examines the effects of demographics on people’s attitudes to environmental protection, but the randomised experiment tests the following four hypotheses. Two images and two forms of text are tested to compare the relative merits of each form of communication. Both positive and negative narrative (regarding economic development and environmental protection) are included to provide balance by viewing the problem from both directions. Non-directional hypotheses are used because relationships in either direction are plausible, and effects can be in unanticipated directions in some contexts and for some people. For example, positive narrative aimed at providing benefits from deforestation can inflame people who are pro-environmental, causing them to be more pro-environmental.
H1: Photos of environmental damage from economic development changes attitudes to environmental protection.
H2: A map showing the extent of economic development and environmental damage changes attitudes to environmental protection.
H3: Positive narrative concerning benefits from economic development causing environmental damage changes attitudes to environmental protection.
H4: Negative narrative concerning environmental damage from economic development changes attitudes to environmental protection.
The possibility that the result for each hypothesis may depend on another randomised variable (for example, the effect of the photos depends on whether it is accompanied by the negative narrative) or demographics (for example, the effect of the photos depends on a respondents age) are considered by testing for interactions.
The above hypotheses can be viewed as effects from providing information, so an important demographic is prior knowledge of the relevant facts and issues. This is particularly relevant in the presence of misinformation, disinformation, and lack of knowledge created when interested parties have incentives to promote different positions (including ignorance) for financial or other personal benefit. This interaction between the information provided in the above hypotheses and the prior knowledge (demographic) of the respondent is explicitly considered with the following hypothesis.
H5: Results for hypotheses H1 to H4 are moderated by the respondent’s prior knowledge concerning environmental damage from economic development.
Materials and methods
Respondents for an online survey were recruited from Researchify, which is a company that provides payment to panels of respondents to complete online surveys. The survey was advertised as research about the Darling Range as a recreational destination. Elements of the methods are summarised in the following sections, with variable names in italics.
Study site: the Darling Range of Western Australia (WA)
The study site is the Darling Range, located from the eastern suburbs of Perth (the capital city and major population centre of Western Australia (WA)) and extending south hundreds of kilometres. While extensive logging of the forests since European settlement almost 200 years ago has transformed the natural environment, complete clear-felling was rarer with logging typically extracting only high-quality timber along tracks created to transport logs (Wardell-Johnson et al. 2019). Bauxite mining started approximately 50 years ago and is generally within a region known as the northern jarrah forest (named after the jarrah tree (Eucalyptus marginata), which is only found locally).
Bauxite mining requires clearing all vegetation and topsoil first, but vegetation is typically replanted after mining. Being close to Perth, most Western Australians have some familiarity with the jarrah forests for recreation, and mining is an employer and contributor to the local economy. A map of the area is in Fig. 1. While scientific evidence previously suggested this biodiverse forest could be restored after bauxite mining, the recent scientific opinion applying international standards for ecological restoration of mine sites result in only two stars out of five (Campbell et al. 2024). For example, some plants (e.g., Xanthorrhoea or Balga or grass tree) grow at most a few centimetres per year and an area planted 20 years ago will have most trees of the same size (and none larger). Bauxite mining and the processing of bauxite into alumina is polluting and requires considerable fresh water, a limited resource in WA.
The seven pages of a hypothetical trip corresponding to all four experimental treatments being present. Copyright information (not shown during survey). Map: 2024 Google, Inc. Map data: GoogleLandsat/CopernicusData SIO, NOAA, U.S. Navy, NGA, GEBCOData LDEO-Columbia, NSF, NOAA. Annotation and legend added by the author. Photographs: Ross Taplin.
Bauxite mining also provides significant economic activity. This generates income for government via royalties and employment for residents. Thus, not only do individuals employed in bauxite mining benefit, everyone in WA benefits from the services that governments provide with revenue generated from bauxite mining. Furthermore, there are restrictions on mining activity (such as embargoes on mining areas of very high ecological value) and the regenerated forest is enjoyed by some residents (e.g. bushwalking and mountain biking trails).
Randomised experimental treatments
A 24 experimental design was adopted with four treatment variables embedded into a hypothetical trip that respondents experienced within the questionnaire (see below). These experimental treatments (Table 1) took values of 1 (shown during the trip) or 0 (not shown during the trip) for four variables: (1) photos, which are images of mining with vegetation removed; (2) map, which shows current and historical mining activity locations; (3) N+, which is a positive narrative about bauxite mining; and (4) N−, which is a negative narrative about bauxite mining. Fig. 1 shows the hypothetical trip when all four experimental variables equal 1 (present). The other 15 of the 24 = 16 experimental treatments are formed by removing combinations of the four treatment variables: by substituting the two photos showing mining with pictures of natural forest (like the first and fourth photos); by removing the map at the end of the hypothetical trip; by removing the positive narrative; and by removing the negative narrative (see Supplementary material for the trip when all four treatments are absent). Respondents were randomly allocated to receive one of the 16 experimental treatments. This randomisation provides stronger evidence of causal conclusions (Ramsey and Schafer 2013).
| Variable | Variable = 1 (present) | Variable = 0 (absent) | |
|---|---|---|---|
| Photos | The third and fifth photo (of 5) shown in the hypothetical trip show evidence of bauxite mining | No photographs show evidence of mining | |
| Map | Map showing the extent of current and historical/regenerated bauxite mining is shown | Map is not shown | |
| N+ | A positive Narrative is shown: ‘Bauxite mining provides jobs for thousands of workers and income to government (eg. royalties). This helps provide essential services for everyone.’ | The positive narrative is not shown | |
| N− | A negative Narrative is shown: ‘Bauxite mining is polluting, places stress on fresh water supplies, and regenerated forests are ecologically different to the original forest.’ | The negative narrative is not shown. |
Since images of forest and mining can vary, three replicates of the photos were used. Each replicate had similar but different images of forests and mining; for example, the fourth photo of Fig. 1 has a young tree in the foreground, but the other replicates had trees in the middle or background. Results were not significantly different for the replicates so are not discussed further. These replicates add generalisability to the results since these results do not depend significantly on the particular photos used (only whether they show mining or not).
Questionnaire
Upon entering the questionnaire, respondents had to provide consent to participate (Curtin University ethics approved: HREC number HRE2024-0050). They then answered demographic questions (gender and education were measured at the end of the questionnaire since these were not needed earlier and are unlikely to be influenced by the content of the survey). Attention check questions were also used to remove inattentive respondents (see Supplementary material). Respondents were then shown a short advertisement about recreational activities such as walk trails in the Darling Range (see Supplementary material) before being shown one of the 16 hypothetical trips (see previous section). Thus respondents were presented with a plausible introduction to the hypothetical trip and the experimental treatments, the impacts of which are discussed in the limitations section below.
The dependent variables concerning environmental attitudes (for the hypotheses) were deliberately chosen to provide balance by asking for attitudes towards both environmental protection and economic development (that has a side-effect of environmental destruction). These were measured with the dependent variables Cpreserve (‘Did seeing this hypothetical trip make you more or less likely to believe the jarrah forest of the Darling Range is worth preserving’) and Cmining (‘Did seeing this hypothetical trip make you more or less likely to believe mining the Darling Range is worthwhile’). Both questions were measured on a 5-point Likert scale with end points of −2 (much less likely) and 2 (much more likely) with a midpoint of 0 (made no difference). Measuring this change in attitude due to the hypothetical trip provides superior measurement of effects due to the treatments.
As a further methodological improvement, respondents were asked to consider both environmental protection and economic development simultaneously by placing themselves on a spectrum between these extremes. This third dependent variable (EnvDev) was measured with ‘If choosing between economic development (jobs and prosperity) and environmental protection, which do you consider more important now?’, measured with a slider (201 points from −1 (economic development) to 1 (environmental protection). This may remove bias created from the fact that both protecting forests and economic development are generally accepted as socially desirable (all other things being equal), but often in practice a choice between the two is necessary. This environment-development spectrum was also measured at the beginning of the survey as a control variable (control).
Past studies have used a dichotomous variable to capture this tension between environmental protection and economic development. For example, a World Health Survey question asks respondents to choose between two statements ‘Protecting the environment should be given priority, even if it causes slower economic growth and some loss of jobs’ and ‘Economic growth and creating jobs should be the top priority, even if the environment suffers to some extent’ (Haerper et al. 2022). This question has been used in academic (non-randomised) research such as Holum and Jakobsen (2024) to establish citizen preferences to protect the environment over economic development is correlated with government spending on environmental protection across 27 countries. This question has also been criticised as a false dichotomy because it is possible to have both, such as with the renewable energy industry (Oreskes 2024). That is, it is not always a choice between environmental protection and economic development, sometimes both are achievable.
Despite this criticism, the environmental-development spectrum question is used for several reasons. First, a slider is used rather than a dichotomy so respondents can choose a position between the two extremes. Second, while it is true that both environmental protection and economic prosperity can be possible, sometimes it is not. Bauxite mining requires removal of all vegetation; it is impossible to leave the forests intact and extract the bauxite. So, bauxite mining is an example where a choice is required. Third, readers who do not accept these arguments can ignore results using EnvDev and concentrate on the results from Cpreserve and Cmining.
One measure of technical knowledge was whether respondents were aware of the extent of mining activity in the Darling Range. This was measured when respondents saw the map (see Fig. 1), which was during the hypothetical trip when Map = 1 and after measurement of the dependent variables when Map = 0. This was measured by asking respondents ‘Does this map show more or less mining than you thought was the case before this survey?’ with possible responses of the ‘map has more mining than I thought’; ‘map is similar to what I thought’, ‘map has less mining than I thought’, and ‘I have no idea’ (corresponding to the variable accurate equalling 0, 1, 2, or 3 respectively).
For the other measure of technical knowledge, two pairs of pictures of jarrah forest were shown near the end of the questionnaire (see Fig. 2). Within each pair one photo was taken by the author at a site known to be regenerated about 20 years ago (the other site known to never have been mined). For each pair, respondents were asked ‘Which would you prefer to visit?’ and, to measure technical knowledge, ‘which is most pristine (not regenerated/replanted)?’. Respondents could select either picture or respond they are similar, and in the case of the second question (technical knowledge) an option ‘I do not know’ was also included. Within each pair, the pictures were presented in random order to reduce order effects. In the first pair, one picture had a Xanthorrhoea (grasstree or balga) clearly visible in the foreground. In the second pair one photo included a large tree. These characteristics (Xanthorrhoea and larger tree) would not be present in a young forest regenerated after bauxite mining. Pictures after bauxite mining also showed many trees, but only small trees of similar size consistent with a forest regenerating after bauxite mining.
Respondents
Respondents were immediately excluded from the survey if in the first three questions they refused consent to participate (ethics requirement), were under 18 years old, or resided outside WA. Respondents were also removed during the survey for failing an attention check question or if they withdrew from the survey (see Supplementary material on data quality).
Respondent demographics of the N = 1022 respondents included for analysis (after exclusions) are provided in Table 2 (see the footnote for definitions). Characteristics worth noting include: wealth, which was generally low with almost half (44%) indicating they perceived they could barely pay essentials; 66% of the respondents are female and younger and more educated respondents are slightly over-represented (see statistical analysis section below); most of the Perth suburbs is covered by location = 2 (5–30 kms from the Darling Range); and few respondents (8%) believe the original ecosystem can be restored after bauxite mining but 26% have no idea. Respondents also tend to place themselves further towards environmental protection than economic development on a spectrum between these two extremes (control). Finally, the high median of 0.57 for diverse indicates most respondents agree the forest is ecologically diverse and the low median of experience indicates most respondents disagree they have extensive experience being in the forests.
| Variable | Value | |||||||
|---|---|---|---|---|---|---|---|---|
| 0 | 1 | 2 | 3 | 4 | 5 | 6 | ||
| wealth | 16% | 44% | 22% | 16% | 3% | |||
| age | 20% | 22% | 22% | 13% | 11% | 10% | 3% | |
| female | 34% | 66% | ||||||
| location | 6% | 15% | 58% | 21% | ||||
| education | 17% | 41% | 42% | |||||
| restore | 34% | 32% | 8% | 26% | ||||
| Minimum | Lower quartile | Median | Mean | Upper quartile | Maximum | ||
|---|---|---|---|---|---|---|---|
| control | −1 | −0.03 | 0.30 | 0.27 | 0.66 | 1 | |
| experience | −1 | −0.85 | −0.39 | −0.31 | 0.09 | 1 | |
| diverse | −1 | 0.19 | 0.57 | 0.52 | 0.87 | 1 |
Variables were anchored/defined as follow:
wealth: ‘Consider the total income, savings and expenses of your family living in your home. What best describes your household’s current financial situation?’ was defined as: (0) struggle to pay for essentials (food, rent/mortgage, etc.); (1) can pay essentials, but not much left over; (2) can pay essentials and some more, like travel for a two week holiday each year; (3) can afford most things for a comfortable lifestyle; (4) can afford almost anything I want.
age (in years): (0) 18–25; (1) 26–35; (2) 36–45; (3) 46–55; (4) 56–65; (5) 66–75; (6) 76+.
female: (0) male; (1) female.
location: ‘Which best describes your usual place of residence?’ was defined with (0) within the hills of the Darling Range: (1) within 5 kms of the Darling Range (e.g. Midland, Armadale); (2) from 5 to 30 kms from the Darling Range; (3) at least 30 kms from the Darling Range (so not in Perth suburbs).
education: (0) less than year 12 high school; (1) year 12 high school; (2) university degree.
restore: ‘After bauxite mining in the Darling Range, can the original ecosystems be restored?’ was defined with (0) no, the ecosystems will be completely different; (1) partly; (2) yes, completely; (3) I have no idea. Variables restoreN and restoreNI were used (see statistical analysis section).
Slider (201 point) scales from (−1) strongly disagree to (1) strongly agree were used for the following three questions:
control: ‘If choosing between economic development (jobs and prosperity) and environmental protection, which do you consider more important’.
experience: ‘I have extensive experience being in the forests of the Darling Range’; or
diverse: ‘The native jarrah forest in the Darling Range is ecologically diverse’.
Statistical analysis
The ability of respondents to distinguish between the pairs of pictures of forest was summarised with Fscore (= 0, 1 or 2), defined as the number of pairs for which they correctly identified which was more pristine. Frequencies of Fscore are presented together with the frequencies of the four possible responses for the accuracy of the map. Relationships between demographics and Fscore (multiple regression) and knowing the map is accurate (logistic regression) are explored.
The hypotheses H1 to H4 are tested using multiple regression, with dependent variables Cpreserve, Cmining, and EnvDev and independent variables corresponding to the four randomised treatments (Photos, Map, N+, and N−) as well as demographic variables, including diverse and restore capturing generic knowledge and accurate and Fscore capturing technical knowledge. Two regression variables were used for restore: restoreN equals restore but with those having no idea (restore = 3) assigned the midpoint value of 1 and a 0/1 dummy variable restoreNI equal to 1 when restore = 3 (= 0 otherwise). Relative to being knowledgeable about the map (accurate = 1), three 0/1 dummy variables accMore, accLess and accUnsure were used to capture whether respondents thought the map showed more mining, less mining or they did not know (accurate = 0, 2, and 3 respectively).
For each dependent variable, regressions were also performed including interaction effects between the randomised experimental variables and each other and with demographic variables. This tested whether results for hypotheses H1 to H4 depended on other variables. For hypothesis H5, interactions between Photos, Map, N+, and N− and the technical measures of knowledge (accurate and Fscore) and the generic measures of knowledge (restore and diverse) are included in regressions. Interactions between demographics and randomised treatments are investigated to describe how the four hypothesised causal effects differ between subgroups of the population rather than incorrectly implying the demographic has a causal effect (Fesenfeld et al. 2024).
The variables exhibit low correlations with each other; the following exceed 0.2 in magnitude. Experience is negatively correlated with location (r = −0.24) and restoreNI (r = −0.22), so respondents with less experience in the forests of the Darling Range tend to live further from the Darling Range and are more likely to have no idea whether the ecosystems can be restored. Diverse is positively correlated with control (r = 0.25), so respondents who believe the forest is diverse place themselves further towards environmental protection than economic development. Importantly, all demographics have negligible correlations (|r| < 0.04, P > 0.2) with the four key independent variables Photo, Map, N+, and N− since these are randomly allocated to respondents.
The high (66%) proportion of females was expected prior to data collection as it would be difficult to obtain the requested sample size of males within Western Australia. While this suggests overall results might be dominated by females, this was mitigated by including female as a control variable and testing whether the effect of the four experimental treatments depend on this variable. The same approach was taken for all demographics variables.
Results
Results for the technical knowledge of respondents are presented first (for comparison with the generic knowledge, diverse and restore are in Table 2), followed by the causal effects of the randomised treatments and then how these causal effects may depend on other variables.
Knowledge of bauxite mining in the jarrah forest of the Darling Range (RQ1)
Regarding their perceptions of the accuracy of the map, 57% of respondents thought the map showed more mining activity than they thought was the case, 26% thought it was similar and 16% thought it showed less (7% indicated they had no idea). These results do not depend significantly on any of the randomised treatments, including whether they saw the negative narrative that stated regenerated forests after bauxite mining are ecologically different to the original forest. Hence only about one quarter of the respondents possessed the technical knowledge that the map was accurate. Believing the map is accurate (accurate = 1) is significantly higher when age (P = 0.009) and education (P = 0.045) are lower and when experience (P = 0.002) and restoreN (P < 0.001) are higher (see Supplementary material).
The significant relationship between accurate and restore (Fig. 3) is largely due to respondents who are more knowledgeable that the ecosystems cannot be restored after bauxite mining (restore = 0) tend to be less knowledgeable about the extent of mining and more likely to believe ethe map shows more mining activity than they thought was the case.
Distribution of knowledge about the extent of mining activity (accurate) conditional on the respondents’ knowledge of whether the forest can be restored after mining (restore).
When asked which of two images (Fig. 2) was most pristine (not regenerated/replanted), more respondents indicated the incorrect image than the correct image (relatively few respondents indicated they did not know or they were similar, Table 3). Thus, most respondents thought they knew the one was more pristine, but most of these respondents were incorrect. The correct image for Pair 1 included a Xanthorrhoea grass tree, but most respondents were apparently unaware these grow too slowly to be grown after bauxite mining. In contrast, 40% of respondents indicated they preferred to visit the image with the Xanthorrhoea and only 36% the other image. So, respondents tended to prefer to visit the site with Xanthorrhoea even though they tended to think it was less pristine.
| Correct | Similar | Incorrect | Don’t know | P-value | ||
|---|---|---|---|---|---|---|
| Pair 1 | 30% | 9% | 50% | 12% | 0.000 | |
| Pair 2 | 37% | 13% | 41% | 10% | 0.094 |
P-tests of the null hypothesis that respondents selecting either the correct or incorrect image are randomly guessing.
Combining results from these two questions to calculate Fscore, 54%, 26% and 20% of respondents provided correct answers to 0, 1, and 2 of these questions. Fscore is significantly higher for respondents with higher age (P < 0.001), diverse (P = 0.014), and experience P = 0.039) but not significantly related to other demographics or the randomised treatments (see Supplementary material). The strongest relationship is with age, with the average Fscore about 0.5 for the youngest age group and almost 1 (double) for the oldest age group.
Regarding research question RQ1, both these results suggest Western Australians are mostly unaware of how extensive the Darling Range has been mined for bauxite and are unable to correctly identify images of forest that have been replanted after mining. Thus, they have low technical knowledge, in contrast to their high generic knowledge (restore and diversity in Table 2).
Attitudes to the environment and mining (RQ2)
First, overall descriptive statistics indicate the hypothetical trip (on average) made most respondents much more likely to believe the jarrah forest in the Darling Range is worth preserving and less likely to believe mining the Darling Range is worthwhile (Table 4).
| −2 | −1 | 0 | 1 | 2 | Mean | ||
|---|---|---|---|---|---|---|---|
| Cpreserve | 1% | 2% | 15% | 26% | 56% | 1.35 | |
| Cmining | 20% | 22% | 34% | 14% | 10% | −0.28 |
−1 = much less likely; 0 = made no difference; 2 = much more likely.
Overall responses to the slider between economic development and environmental protection (EnvDev) also suggest the hypothetical trips on average increased attitudes to protecting the environment (Table 5). On the scale from −1 (economic development) to 1 (environmental protection), the mean is 0.42 and over 75% give a score above the midpoint of 0 (first quartile is 0.08). Furthermore, comparing with the responses to the same question prior to the hypothetical trip (control), EnvDev is higher by on average 0.15 (Table 5).
| Min | Q1 | Median | Mean | Q3 | Max | ||
|---|---|---|---|---|---|---|---|
| EnvDev | −1 | 0.08 | 0.52 | 0.42 | 0.84 | 1 | |
| EnvDev-control | −2 | −0.06 | 0.08 | 0.15 | 0.38 | 2 |
Reasons for these overall trends are elaborated on in the Discussion section, but first the effects of the four randomised treatments within the hypothetical trip are discussed because the randomisation provides stronger evidence for conclusions of causality.
Causal effects of randomised treatments on environmental attitudes (H1−H4)
Each of the three dependent variables (Cpreserve, Cmining, and EnvDev) capturing attitudes about environmental protection were significantly influenced by at least one of the four randomised variables (Table 6). The negative narrative N− significantly (P = 0.035) increased respondents desire to protect the forest (Cpreserve) but Photos (P = 0.854), Map (P = 0.928) and N+ (P = 0.889) had statistically insignificant effects. The worth of mining (Cmining) was significantly increased by Map (P = 0.039) and the positive narrative N+ (P < 0.001) while EnvDev was significantly (P < 0.001) decreased by N+. Photos did not significantly impact any of the three dependent variables. Thus hypothesis H1 is not supported (however, see below for a significant interaction effect). There is evidence in support of hypotheses H2, H3 and H4 (although results are not always consistent across all three dependent variables).
| Variable | Cpreserve | Cmining | EnvDev | |||||||
|---|---|---|---|---|---|---|---|---|---|---|
| B | s.e. | P-value | B | s.e. | P-value | B | s.e. | P-value | ||
| Photos | 0.01 | 0.05 | 0.854 | −0.09 | 0.07 | 0.206 | 0.02 | 0.02 | 0.464 | |
| Map | 0.00 | 0.05 | 0.928 | 0.15 | 0.07 | 0.039* | −0.04 | 0.02 | 0.111 | |
| N+ | 0.01 | 0.05 | 0.889 | 0.30 | 0.07 | 0.000*** | −0.08 | 0.02 | 0.000*** | |
| N− | 0.10 | 0.05 | 0.035* | −0.07 | 0.07 | 0.304 | 0.04 | 0.02 | 0.113 | |
| control | 0.07 | 0.06 | 0.205 | −0.07 | 0.08 | 0.414 | 0.48 | 0.03 | 0.000*** | |
| control 2 | 0.25 | 0.08 | 0.002** | −0.30 | 0.12 | 0.010* | 0.04 | 0.04 | 0.256 | |
| wealth | 0.01 | 0.02 | 0.788 | −0.03 | 0.04 | 0.387 | 0.01 | 0.01 | 0.351 | |
| age | −0.01 | 0.02 | 0.613 | 0.00 | 0.02 | 0.932 | −0.03 | 0.01 | 0.001*** | |
| female | 0.28 | 0.06 | 0.000*** | −0.13 | 0.08 | 0.095 | 0.04 | 0.03 | 0.112 | |
| location | 0.03 | 0.03 | 0.393 | 0.02 | 0.05 | 0.622 | 0.00 | 0.02 | 0.769 | |
| education | 0.00 | 0.04 | 0.970 | −0.01 | 0.05 | 0.811 | −0.03 | 0.02 | 0.132 | |
| restoreN | −0.12 | 0.05 | 0.008** | 0.48 | 0.07 | 0.000*** | −0.08 | 0.02 | 0.000*** | |
| restoreNI | 0.04 | 0.06 | 0.516 | −0.28 | 0.09 | 0.003** | 0.06 | 0.03 | 0.039* | |
| experience | 0.03 | 0.05 | 0.558 | 0.13 | 0.07 | 0.059 | 0.01 | 0.02 | 0.516 | |
| diverse | 0.36 | 0.07 | 0.000*** | −0.11 | 0.10 | 0.258 | 0.10 | 0.03 | 0.002** | |
| accMore | 0.45 | 0.07 | 0.000*** | −0.35 | 0.09 | 0.000*** | 0.09 | 0.03 | 0.004** | |
| accLess | 0.27 | 0.08 | 0.001** | 0.19 | 0.12 | 0.125 | −0.10 | 0.04 | 0.014* | |
| accUnsure | −0.09 | 0.11 | 0.413 | 0.07 | 0.16 | 0.663 | −0.11 | 0.05 | 0.036* | |
| Fscore | 0.02 | 0.03 | 0.502 | −0.03 | 0.05 | 0.447 | 0.02 | 0.02 | 0.145 | |
| (Intercept) | 0.54 | 0.13 | 0.000*** | −0.17 | 0.19 | 0.000*** | 0.30 | 0.06 | 0.000*** | |
*P < 0.05; **P < 0.01; ***P < 0.001.
Several non-randomised variables are also significant (Table 6) and the control variable capturing initial attitudes on the environment-development spectrum before their hypothetical trip (control) is a strong predictor of all three dependent variables. This relationship is non-linear for Cpreserve and Cmining with a stronger relationship when control is higher (e.g. changing control from −1 to 0 has a smaller impact on Cpreserve and Cmining than changing control from 0 to 1). Cpreserve is significantly (P < 0.001) higher for females but this relationship is not statistically significant for Cmining (P = 0.095) and EnvDev (P = 0.112). Thus, the descriptive statistics for Cpreserve (Table 4) are biased relative to the general population as males are significantly less supportive, but only by 0.28 on average. The distribution of responses for Cpreserve (Table 4) for males are 1%, 3%, 22%, 30%, and 44% respectively, so 74% of males (compared to 87% of females) report an increase in belief the jarrah forest is worth preserving. Respondents who belief the ecosystems can be restored after mining (restoreN) also report significantly lower Cpreserve (P = 0.008), higher Cmining (P < 0.001), and lower EnvDev (P < 0.001).
For the technical measures of knowledge of bauxite mining in the jarrah forests, identifying which images of forests were or were not regenerated after bauxite mining (Fscore) was not significant but knowledge of the extent of mining was significant for all three attitudes. Compared to 20% of respondents who thought the map was accurate, the 57% of respondents who indicated the map showed more mining activity than they believed was the case (accMore) had significantly higher Cpreserve (P < 0.001), lower Cmining (P < 0.001) and higher EnvDev (P = 0.004). The respondents (16%) who indicated the map showed less mining activity than they believed was the case (accLess) had significantly (P = 0.014) lower EnvDev and insignificantly (P = 0.125) higher Cmining. For Cpreserve, attitudes to protect the forest was significantly higher for respondents who believe the map showed more (accMore; P < 0.001) and for respondents who thought it showed less (accLess; P = 0.001) mining than they previously believed was the case (compared to those who thought it was similar). This inconsistency is considered in the ‘Discussion’ section.
Interactions and moderation of treatments (H5)
First, none of the four randomised treatment variables had significant interactions with other treatment variables. Hence the effects of the randomised treatments can be considered in isolation (i.e. additive). Most demographic variables did not have significant interactions with the four treatment variables either. For example, interactions between female and each of the four experimental variables Photos, Map, N+, and N− were insignificant for each of the three dependent variables. This result is important because it suggests the over-sampling of females has not introduced a bias in the causal results of the treatments.
Results provide little support for H5; evidence the hypothesised effect in H1 to H4 depend on the technical or generic knowledge of respondents is weak. Hypothesised interactions involving technical knowledge (accurate and Fscore) are all insignificant except Photos*Fscore (P = 0.030) for EnvDev, where the effect of seeing the photos was to increase EnvDev by 0.06 for respondents with Fscore = 0 (low knowledge) and decrease it by 0.07 when Fscore = 2 (high knowledge). This coefficient of 0.06 (P = 0.049) applies to 54% of the sample who could not correctly identify either pair of regenerated/pristine forest. Note earlier (Table 6) that the overall effect (averaged across all respondents) on EnvDev of the photos of mining was insignificant (B = 0.02, P = 0.464). This significant interaction suggests the effect of the photos of mining depends (and is in different directions) on the ability of the respondent to differentiate between images of forest with and without bauxite mining. The effect of the photos of mining is to make respondents with this knowledge more pro-environmental and respondents without this knowledge more pro-mining on the environmental−development spectrum. This can therefore be interpreted as some support for hypothesis H1 when considering subsets of respondents.
The insignificant (P = 0.752) interaction between Map and accurate for Cmining (Fig. 4) is briefly reported and discussed because both these variables are significant, it was a hypothesised relationship (H5), it is plausible they should display an interaction (if the map has an effect, then this effect might be expected to depend on what new knowledge it provided) and it illustrates important issues when interpretating results. The largest effect of the map on Cmining was for the respondents who indicated the map was similar to what they previously believed (so it was not informing them of facts they were unaware of). This is contrary to expectations if new information would alter attitudes. The lack of a significant interaction may partly be due to lack of power because few respondents thought the map was accurate. However, the perceived accuracy of the map has a strong effect on Cmining even for the respondents who, when Cmining was measured, had not yet seen the map. This provides evidence accurate is measuring characteristics of the respondents unrelated to their reaction to the map (see ‘Discussion’).
Estimated change in belief the forest should be mined (Cmining) depending on whether the map was shown (Map) conditional on perceptions of the map relative to prior belief about the extent of mining (accurate).
For Cmining, there was a significant (P = 0.037) interaction between Map and the generic knowledge variable restore, with the map increasing Cmining by 0.32 for respondents who believe the ecosystems cannot be restored (high generic knowledge) and decreasing Cmining by 0.31 for those who believe it can be fully restored (low generic knowledge). However, contradictorily, respondents unsure about restoration (restoreNI = 1) experience an increase of Cmining by 0.23, similar to those with high knowledge.
The interaction between Photos and the generic knowledge variable diverse was also significant (P = 0.024) for Cmining, with effects varying from B = −0.69 (lowest diverse) to 0.10 (highest diverse), suggesting the photos of mining may cause a decrease in attitudes towards mining for respondents who disagree the forest are diverse (low generic knowledge).
Although not specifically hypothesised, there were a few other significant interaction effects. For Cpreserve, N+ significantly interacted with age (P = 0.038), with the effect of N+ ranging from B = 0.14 (youngest) to −0.22 (oldest). Location significantly interacted with N+ (P = .031) and N− (P = 0.003) for Cpreserve, with coefficients ranging from B = 0.28 and 0.47 (for respondents living in the Darling Range) to −0.14 and −0.28 (for those living over 30 kms away). Other interactions not reported above were insignificant (P > 0.05). Evidence for significant interactions reported above is typically not strong (P > 0.01) so might be better interpreted as exploratory.
Discussion
Regarding the first research question, Western Australians typically agree the jarrah forests are ecologically diverse and understand the ecosystems cannot be fully restored after bauxite mining. However few recognise distinctive features of forest (Xanthorrhoea and large, old trees) that are not present in young forest regenerated after bauxite mining. It could be concluded that people only have superficial knowledge. Agreeing the forest is ecologically diverse is not correlated with recognising biodiversity in images of forests. In a democracy, this result provides assistance to bauxite mining since the general public are unlikely to complain about something they are ignorant about. However, expert opinion is also important, and in this case ecologists argue the forests are different (Campbell et al. 2024). This difference between ecologists and the general public might suggest the need to educate the general public, not only because education is an end in itself but because in democracies the views of the public have consequences. Furthermore, only one quarter of Western Australians thought the extent of mining in the Darling Range was similar to that shown by the map, with the majority thinking the map showed more mining. Most people are unaware of the scale of bauxite mining.
These results suggest people provide socially acceptable answers to generic questions about diversity and restoration but cannot answer simple technical questions, such as what diversity looks like and the extent of mining activity. For example, most respondents agree the forest is biodiverse, but few recognise signs of biodiversity in images of forest. Thus, technical measurements of knowledge may be preferable to measures that may be mis-interpreted or subject to social-desirability bias.
The regression results consistently show respondents become more pro-environmental and less pro-mining when they believe the map shows more rather than less mining than their prior belief, but with one exception; respondents are more in favour of protecting the forest even if they indicate the map shows less mining (as well as more mining) than their prior belief. Ignoring this exception, it appears responses regarding the accuracy of the map convey information about the environmental attitudes of the respondent; people who think there is more mining than they thought was the case tend to be more pro-environmental and less pro-mining. It is noted that this trend does not differ significantly depending on whether respondents see the map prior to answering the three attitude questions, so the map is not causing this effect. However, the exception provides insights concerning the measurement of environmental attitudes; people who believe the map shows less mining than they thought was the case may be relatively pro-environmental, but not at the expense of economic prosperity. For these respondents, this difference in results from support to protect the forests (Cpreserve) and responses to the environmental-development scale (EnvDev) may be explained by the presence of different types of environmentalists, those who desire protection at all costs and others who desire protection but not at the expense of economic prosperity. This may also manifest in responses to the environmental-development scale itself, with respondents choosing the most extreme pro-environmental response on this scale potentially implying they prefer environmental protection regardless of the opportunity cost from lost economic development.
These results concerning having technical knowledge do depend on demographics of the respondent in expected ways, such as people with more experience in the jarrah forest being significantly more knowledgeable about the extent of mining and the different characteristics of forest regenerated after bauxite mining. However, while older people are more likely to correctly distinguish forests after mining from those without mining (Fscore), they are less likely to think the map was accurate (accurate = 1). They are more aware of bauxite mining and its impact, but less aware of how it has expanded over time (or find it unacceptable). While the positive relationship between people believing the forest is diverse and recognising differences between regenerated forest and more pristine forest provides evidence of validity, this must be interpreted in the context that most people believe the forests are diverse but cannot correctly distinguish signs of diversity.
Regarding the second research question, the hypothetical trip increased attitudes towards protecting forests and decreased attitudes to bauxite mining (overall, ignoring which hypothetical trip they experienced). This result is likely to be due to the information common to all hypothetical trips, such as the images of forests and details about how bauxite mining requires ‘clearing all the vegetation’, even though they are informed the vegetation is ‘regenerated with new vegetation’. The context of a recreational trip may evoke ideas of pleasure away from the pressures of life, in contrast to a state of mind concerned with employment and cost of living pressures. Hence recreation may be a context useful for improving attitudes towards environmental protection, but economic contexts such as employment more successful to achieve the opposite. However, the lack of randomisation (since the survey presented a recreational trip as context for all respondents) of these features makes conclusions of causality less convincing.
Causal effects of the four randomised treatments corresponding to the hypotheses can be made with greater confidence, although these should be interpreted in light of the overall increase in attitudes to environment protection. For hypothesis H1 there is insignificant evidence images of forests cleared for mining changes attitudes to environmental protection and bauxite mining. The only provisor to this is these photos of deforestation may lower belief that mining bauxite is worthwhile for the small number of respondents who do not agree the jarrah forest is diverse (low knowledge). This does not appear to be due to other observed demographics, such as the education of respondents because these interactions were not statistically significant.
For hypothesis H2 there is evidence the map increased attitudes towards mining overall, but the moderation of this effect suggests this effect only occurs for two types of respondents; the 34% who believe the original ecosystems will be completely different after bauxite mining, and the 26% indicating they had no idea if the ecosystems can be restored (restore). Note this question was the first time bauxite mining was mentioned in the survey, so responses were not impacted by information such as the narrative within the hypothetical trip. It is possible the map showing large areas of historical and regenerated areas in green (a colour associated with lush vegetation or with positivity, such as green traffic lights) was incorrectly interpreted by respondents as information the forests have been restored successfully. This was not the intention and is contrary to scientific evidence (Campbell et al. 2024) and therefore reinforces the first challenge relating to ‘facts’ (Vella et al. 2021). Knowledge does matter, because it is perceptions of ‘facts’ that matter to attitudes. However, a counter-argument is that the negative narrative, where respondents were told the forest was completely different after regeneration, did not significantly moderate the effect of the map. New information may not always change a respondent’s preconceived knowledge. Interpretation of the map is discussed further under future research below.
For hypothesis H3, the positive narrative increased belief mining is worthwhile and caused a corresponding shift on the environment-development spectrum (or more accurately reduced the overall effect of the hypothetical trip to become more pro-environmental). This result may be unsurprising because jobs and economic prosperity are important. Although this pro-mining narrative did not significantly impact attitudes to protect the forest overall, it did lower this attitude for older respondents and raise it for younger respondents. Thus, the pro-mining narrative can increase attitudes towards protecting the forest amongst some respondents, perhaps because they reject and are offended by the argument.
For hypothesis H4, the negative narrative that highlighted the disadvantages of mining had an expected increase in attitudes for environmental protection, and this effect was higher for respondents living closer to the Darling Range but negative for those living far away. This provides evidence that attitudes towards environmental protection is easier to increase when a local environment is under consideration.
The importance of randomised experiments is evident from the significant effect of perceptions about the accuracy of the map (accurate) and effect of seeing the map (Map), but not their interaction (Fig. 4). Perceptions about the map have similar effects on support for mining regardless of whether they have previously seen the map (it is possible a few respondents were aware of the extent of mining from other sources, but none had seen this map that was specifically created for this study). The conclusion from this is that it is not seeing a map that differs from their prior beliefs that impacts support for mining (as might incorrectly be concluded from data if all respondents saw the map before measurement of the dependent variables), but some other unobserved characteristic of the respondents that is related to this perception about the map. While it is possible to speculate on what this unobserved variable is, it is not one of the observed variables (such as experience in the forest or how pro-environmental they are) as these were controlled for in regressions. This suggests the above literature without randomised treatments should be interpreted cautiously, especially if the aim is to infer which interventions will change attitudes to environmental protection.
Implications
Practical implications include possible strategies to change attitudes to deforestation, and the role of educating people with related information. For example, negative images of deforestation due to a mine site may distress people without changing attitudes, but the negative narrative about bauxite mining did change attitudes, especially for local residents. This may have broader implications, including for changing attitudes to global climate change. First, narrative can be at least as influential as images and should not be discounted in future research, in contrast to previous findings by Bolsen et al. (2019). Second, the use of local environments may have more impact. The use of these, even for attitudes to global issues such as climate change, may be worth investigating in future research. Hart (2011) found thematic information was superior to episodic information, but it may be the use of a local issue is important to change attitudes, even for the environmental-development scale (EnvDev) which was not framed in terms of a specific local case. Changing attitudes to a local environment may be a pre-cursor to changing attitudes to global environmental issues.
While the randomised treatments suggest it may be easier to change attitudes towards development and away from environmental protection, the hypothetical trip overall increased attitudes towards environmental protection and away from development. This is important because common elements of the hypothetical trip provided educational information. Hence the treatment effects to increase attitudes towards development and away from environmental protection may be more about minimising effects of information in the opposite direction. Future research could investigate this further; with literature only investigating strategies to move attitudes in one direction potentially limiting and biased. The context may also be important, with positive contexts such as recreation being more effective than negative contexts of a future under the effects of climate change.
Bakaki and Bernauer (2016) ask what rich nations can do to help prevent deforestation in Brazil after concluding locals are too dependent on economic considerations such as farming. While this paper does not directly answer this question, it does provide related insights regarding what residents of a richer country think about deforestation in their own lands. Perhaps the situation in the relatively rich WA is not so different to that in Brazil; locals who understand the economic benefits of activity that requires deforestation are less likely to support protection of the forests. Furthermore, benefits gained from randomised messages such as the negative effects of deforestation are largely negated when respondents are asked to directly respond to an environmental-development spectrum, but positive narrative of the benefits of mining persist regardless of whether pro-mining or the environmental−development spectrum is used. Economics matters. When immediate economic benefits require deforestation, a forest without any obvious intrinsic economic value is more likely to be sacrificed.
Methodological implications include the need to consider how attitudes to environmental protection are measured. For example, all other things being equal, most respondents believe forests should be protected, but these other things are not equal; protecting forests comes with an economic cost (especially to locals). This is central to attitudes and debates on climate change. Alternative measurements include asking whether they are in favour of development or asking respondents to place their attitude on a spectrum between two extremes.
The need for randomised experiments rather than finding associations between non-randomised variables is also demonstrated, especially for understanding results and conclusions about causal effects (which are important for developing strategies to change attitudes). For example, perceptions about the accuracy of the map are strongly related to attitudes concerning environmental protection and bauxite mining, but it is dangerous to make causal conclusions (for example, educating people with this map does not appear to change attitudes towards environmental protection). While these randomised experiments may be expected and common in scientific research, they are less common but just as important in social science research, such as attitudes to climate change.
Limitations and future research
This study examines attitudes rather than whether actual behaviour can be changed. For example, people may have less positive attitudes to bauxite mining but still consume aluminium (the result of processing bauxite). The initial context of taking a recreational trip to the Darling Range might provide a plausible context for the randomised treatments to be applied, but the pleasurable activity of a recreational trip may have influenced respondents. For example, overall it produced strong increases in attitudes to protect the environment. Indeed, this effect was so strong that it resulted in relatively small variation in Cpreserve (a majority of respondents gave the most positive increase possible in attitude to protect the environment) impacting statistical power for this variable compared to Cmining. Future research might need to consider other contexts and methods.
Other randomised treatments might have more impact, especially if more detail is provided. For example, the negative narrative stated the regenerated forests are ecologically different to the original forest, but not how. The short quiz using images of forest after bauxite mining and without mining (Fig. 2) might be fun and provide knowledge quickly, and with hindsight these could have been randomised to appear at either the beginning or the end of the survey to enable testing of causal effects from this quiz. Further research might investigate the effect of this treatment on people and compare the effectiveness of such gamification with simply providing equivalent information about biodiversity. Scientific literature on the extent to which forests can be restored after bauxite mining (Campbell et al. 2024) may also provide more successful technical knowledge to change attitudes.
A limitation of this study is its focus exclusively on flora, but fauna is also important. Further research might test benefits of considering fauna to influence attitudes to flora. For example, koalas (Phascolarctos cinereus) are a well-known Australian animal in danger of extinction because they feed on only a few kinds of eucalypt trees under threat, so saving the koala could save trees. Fielding et al. (2022) found positive attitudes of community members towards protecting koalas was correlated with demographics such as knowledge about koalas. However, their observational study provides questionable evidence that increasing knowledge of koalas will lead to increased attitudes towards koala protection (the causality may, for example, be that attitudes to protect koalas causes people to obtain knowledge about koalas).
Similarly, protecting the northern jarrah forest is identified as a key action to protect cockatoos such as the Carnaby’s cockatoo (Zanda latirostris) for foraging, breeding and roosting habitat (Department of Parks and Wildlife 2013). By comparing two similar species of black cockatoos (Zanda latirostris and Zanda baudinii), Ainsworth et al. (2016) highlight how social factors (such as different threats to agriculture by different species of black cockatoo) can impact attitudes and scientific effort to protect a species. However, their research emphasises fauna over flora and considers views of experts rather than views and knowledge of the general public. Further research might emphasise flora or give both equal emphasis in recognition of interconnections within the overall ecology. Further research on knowledge and attitudes of the public is also encouraged. The use of randomised experiments to test causal effects of interventions to change public attitudes is also encouraged.
Some treatments in this study have several parts, so it is difficult to determine which part had an effect (or whether these effects cancel each other). For example, the map shows both current and historical mining activity and the negative narrative contains information about pollution, water supplies, and how regenerated forests are ecologically different. Further research is required to determine which components of these treatments cause changes in attitudes. The map showing current (yellow) and historical and regenerated (green) mining activity may have been construed as suggesting the forest can be fully restored so future research might explicitly explore how the map is interpreted and the effect of subtle features of the map such as colours. Alternatively, providing information in Campbell et al. (2024) to respondents might be used to test strategies to change public attitudes.
While the negative narrative increased support to protect the forests, the photos of mining activity did not. It is not clear from this data why this is the case as it is plausible seeing direct evidence of forest destruction might have more impact than less direct information by overhearing conversations. It is possible that once a forest is destroyed public opinion to protect it diminishes because it is no longer considered worth saving. Further research concerning both images versus narrative and the impact of damage versus potential damage is warranted.
While results for the three attitude dependent variables are generally consistent, differences exist. This might be due to the measurement of the environmental-development scale where respondents are made to choose a position between two objectives. This may represent how respondents change their level of support for environmental protection when the dimension of economic development is made explicit. It might also reflect measurement issues with asking people to choose a position between two extremes when there are multiple dimensions to consider. However, criticism can also be applied to the other two attitudes; for example, asking people to select their support for protecting the forest without explicitly including the consequences for economic development demands consideration of only one objective rather than two. Further research is warranted both for the case of bauxite mining in the jarrah forest of the Darling range and also more generally, such as attitudes towards climate change.
Only two pairs of images of forest before/after bauxite mining were used to measure knowledge and further research could benefit by testing more images in case there were characteristics specific to these images (such as the amount of green colour) that affected respondents. Furthermore, although the question asked which was more pristine, so asked for a comparison, a case can be made that none of the forest is pristine due to extensive logging. Alternative wording might be investigated in future research.
Deforestation deserves further research because this paper shows attitudes on this issue can be changed, for better or worse, not only for a specific local forest but also attitudes towards environmental protection more broadly. Limiting deforestation is one of the most promising areas to combat global warming (Lawrence et al. 2022), so research on attitudes to deforestation is arguably just as important as research into attitudes to protecting fauna or the use of fossil fuels. This research should include not only changing attitudes but also teaching knowledge, for example by examining the effectiveness of having students distinguish between the biodiversity of different images of forests.
Conclusion
Attitudes to environmental protection is influenced by several factors, including knowledge of the environment. This paper contributed in two ways: (1) it shows how despite people claiming they know a local forest is biodiverse and its ecosystems cannot be restored after mining, they do not recognise simple differences in images of forest regenerated after bauxite mining and images of forest without mining; and (2) in using a randomised experiment, it also provides evidence for how images and narrative providing information can change attitudes. It made several methodological innovations and considered information that can increase and information that can decrease attitudes towards environmental protection, both of which people are exposed to in practice. This paper adds to the complex and difficult topic of changing attitudes, including the interplay between prior knowledge and new knowledge. Attitudes to deforestation is under-researched and deserves further research.
Data availability
The data that support this study are available in FigShare at 10.6084/m9.figshare.27914964.
Acknowledgements
The map in Fig. 1 was produced by the author in Google Earth Pro using the historical feature to approximate current and historical mining, with assistance from websites including mining companies, environmental groups and the Environmental Protection Authority (EPA) of Western Australia. Other images are from the author’s personal photographs. Emily Taplin and Anwen Taplin provided assistance with proof-reading the questionnaire and manuscript drafts while Jennie Wise provided assistance with background information concerning bauxite mining.
References
Ainsworth GB, Aslin HJ, Weston MA, Garnett S. (2016) Social values and species conservation: the case of Baudin’s and Carnaby’s black-cockatoos. Environmental Conservation 43, 294-305.
| Google Scholar |
Bain PG, Hornsey MJ, Bongiorno R, Jeffries C (2012) Promoting pro-environmental action in climate change deniers. Nature Climate Change 2, 603.
| Crossref | Google Scholar |
Bain PG, Milfont TL, Kashima Y, Bilewicz M, Doron G, Garðarsdóttir RB, et al. (2016) Co-benefits of addressing climate change can motivate action around the world. Nature Climate Change 6, 154-157.
| Crossref | Google Scholar |
Bakaki Z, Bernauer T (2016) Measuring and explaining the willingness to pay for forest conservation: evidence from a survey experiment in Brazil. Environmental Research Letters 11, 114001.
| Crossref | Google Scholar |
Bernauer T, McGrath LF (2016) Simple reframing unlikely to boost public support for climate policy. Nature Climate Change 6, 680-683.
| Crossref | Google Scholar |
Boakes Z, Mahyuni LP, Hall AE, Cvitanovic M, Stafford R (2023) Can coral reef restoration programmes facilitate changes in environmental attitudes? A case study on a rural fisher community in North Bali, Indonesia. Human Ecology 51, 891-905.
| Crossref | Google Scholar |
Bolsen T, Palm R, Kingsland JT (2019) Counteracting climate science politicization with effective frames and imagery. Science Communication 41, 147-171.
| Crossref | Google Scholar |
Campbell T, Dixon KW, Bradshaw SD, Gann GD, Hartley W, Lambers H, Wardell-Johnson G (2024) Standards-based evaluation inform ecological restoration outcomes for a major mining activity in a global biodiversity hotspot. Restoration Ecology 32, e14236.
| Crossref | Google Scholar |
Curnock MI, Marshall NA, Thiault L, Heron SF, Hoey J, Williams G, Taylor B, Pert PL, Goldberg J (2019) Shifts in tourists’ sentiments and climate risk perceptions following mass coral bleaching of the Great Barrier Reef. Nature Climate Change 9, 535-541.
| Crossref | Google Scholar |
Erikson L, Klapwijk MJ (2019) Attitudes towards biodiversity conservation and carbon substitution in forestry: a study of stakeholders in Sweden. Forestry 92, 219-229.
| Google Scholar |
Fesenfeld L, Beiser-McGrath L, Sun Y, Wicki M, Bernauer T (2024) Systematic mapping of climate and environmental framing experiments and re-analysis with computational methods points to omitted interaction bias. PLoS Climate 3, e0000297.
| Crossref | Google Scholar |
Fielding K, Lunney D, Rhodes J, Goldingay R, Hetherington S, Brace A, Vass L, Hopkins M, Swankie L, Garofano N, Goulding W, McAlpine C (2022) What predicts community members’ intentions to take action to protect koalas? Pacific Conservation Biology 29, 26-37.
| Crossref | Google Scholar |
Griscom BW, Adams J, Ellis PW, Houghton RA, Lomax G, Miteva DA, et al. (2017) Natural climate solutions. Proceedings of the National Academy of Sciences of the United States of America 114, 11645-11650.
| Crossref | Google Scholar |
Hart PS (2011) One or many? The influence of episodic and thematic climate change frames on policy preferences and individual behavior change. Science Communication 33, 28-51.
| Crossref | Google Scholar |
Holum ML, Jakobsen TG (2024) Economic growth versus the environment: government spending, trust, and citizen support for environmental protection. Environmental Sociology 10, 420-431.
| Crossref | Google Scholar |
Hung LS, Bayrak MM (2020) Comparing the effects of climate change labelling on reactions of the Taiwanese public. Nature Communications 11, 6052.
| Crossref | Google Scholar |
La V-P, Nguyen M-H, Vong Q-H (2024) Climate change denial theories, skeptical arguments, and the role of science communication. SN Social Sciences 4, 175.
| Crossref | Google Scholar |
Lawrence D, Coe M, Walker W, Verchot L, Vandecar K (2022) The unseen effects of deforestation: biophysical effects on climate. Frontiers in Forests and Global Change 5, 756115.
| Crossref | Google Scholar |
Lockie S, Bartelet HA, Ritchie BW, Demeter C, Taylor B, Sie L (2024) Australians support multi-pronged action to build ecosystem resilience in the Great Barrier Reef. Biological Conservation 299, 110789.
| Crossref | Google Scholar |
Maynard LT, Torchalski JR, Gezon ZJ, Callwood KA, Stamper MA, Schook MW, Martin C (2024) Does active or informative messaging result in greater conservation engagement? Conservation 4, 236-252.
| Crossref | Google Scholar |
Mendy L, Karlson M, Lindvall D (2024) Counteracting climate denial: a systematic review. Public Understanding of Science 33, 504-520.
| Crossref | Google Scholar | PubMed |
Oreskes N (2024) Environmental protection does not kill jobs. Scientific American Magazine 330, 80.
| Crossref | Google Scholar |
Roe S, Streck C, Obersteiner M, Frank S, Griscom B, Drouet L, et al. (2019) Contribution of the land sector to a 1.5°C world. Nature Climate Change 9, 817-828.
| Crossref | Google Scholar |
Steg L, Vlek C (2009) Encouraging pro-environmental behaviour: an integrative review and research agenda. Journal of Environmental Psychology 29, 309-317.
| Crossref | Google Scholar |
Trisos CH, Merow C, Pigot AL (2020) The projected timing of abrupt ecological disruption from climate change. Nature 580, 496-501.
| Crossref | Google Scholar | PubMed |
Vella K, Baresi U, Lockie S, Taylor B (2021) Challenges and opportunities for assisted regional ecosystem adaptation: international experience and implications for adaptation research. PLoS ONE 16, e0257868.
| Crossref | Google Scholar | PubMed |
Wardell-Johnson G, Wardell-Johnson A, Schultz B, Dortch J, Robinson T, Collard L, Calver M. (2019) The contest for the tall forests of south-western Australia and the discourses of advocates. Pacific Conservation Biology 25(1), 50-71.
| Crossref | Google Scholar |
Zeng D, Chen B, Wang J, Innes JL, Lu J, Guo F, Yan Y, Wang G (2023) Evolving environmental awareness and shifts in management priorities: a socioeconomic lens on the min river basin, China. Frontiers in Environmental Science 11, 1257089.
| Crossref | Google Scholar |
