Can combining education and entertainment in video games promote pro-environmental behaviour?

We conduct an online randomized experiment to evaluate the potential of entertaining video games to promote pro-environmental behaviour. Our study exploits the interactive nature of video games and delivers an intervention on sustainable food consumption through a newly designed simple single-player mobile game with an entertaining main story that is not inherently environmental.

We developed multiple versions of the game that all include food choices and share the same core elements — storyline, characters, world, and gameplay. Each version also has the same length (estimated median completion time of about 40 minutes). However, the versions differ in the inclusion of educational content about sustainable food consumption and in the way feedback on food choices and their consequences are presented to the player. We created four versions of the same game: Control, Nature, Social, and Social+Nature, following a 2x2 design.

The Control version offers a purely entertaining experience without any explicit environmental or educational content. The gameplay will still include food choices throughout the story, but after each food choice the PC does not receive any information and feedback. The choices made will also have no impact on the story ending. In contrast, the three treatment versions use the food choices to introduce educational content about sustainable food consumption, which accounts for roughly 15% to 20% of the text. All three treatment versions include the same scientific information, but differ in the way feedback and consequences are presented. Specifically, the PC receives feedback and information about environmental impacts after each in-game food choice based on scientific estimatesThe feedback might come from the in-game environment, showing how it flourishes or deteriorates, from society, where NPCs evaluate the PC’s choices, or from both. Additionally, the consequences of in-game choices are exemplified through different story endings depending on the choices made.

In the Nature (N) treatment, we target the challenge that in real life, people often struggle to mentally connect their small daily choices today with the global future consequences of climate change and environmental degradation, which can feel distant, abstract, and thus hard to imagine vividly. In this game version, feedback on the PC’s food choices is conveyed through visual changes in the in-game environment. Sustainable choices are associated with a green and lush landscape, while unsustainable choices are associated with visible environmental damage, depicted through images in magazine articles. The game culminates with an ending that reflects the PC’s decisions, with a flourishing natural environment in the "good" ending where the PC has made mostly green choices, and noticeable degradation in the "bad" ending, where the majority of choices were unsustainable. By directly linking the player’s choices to environmental outcomes into the game, the Nature treatment aims to create vivid mental imagery and associations between individual decisions and their environmental impact. This could help bringing environmental considerations to the top of their minds also in real-life behaviour.

In the Social (S) version of the game, we build on existing research that demonstrates the power of social norms and peer effects to encourage sustainable choices. One challenge is that individuals may have difficulty envisioning drastic individual and societal changes over time. Fictional stories have been suggested to serve as a “simulation of social worlds”, conveying information, values, and norms through immersive social interactions experiences. Therefore, in this version, feedback is provided through social interactions that reinforce social norms and a green social identity. NPCs actively comment on the PC’s food choices. These dialogues provide a collection of social cues, including descriptive and injunctive norms as well as reputational and status concerns. The game concludes with shifting social norms towards environmentally conscious behaviour, culminating in the PC’s visit to a “popular club” focused on sustainable living. In the "good" ending, if the PC has made mostly green choices, they are invited to join the club; otherwise, in the "bad" ending, they are asked to reconsider their past actions. This might enable players to envision an alternate societal future that is partly shaped by their own actions. By incorporating NPCs who evaluate the PC’s choices, our Social treatment seeks to simulate social dynamics and reinforce new norms of behaviour through immersive social interactions.

The Nature+Social version combines elements from both the Nature and Social versions. To avoid information redundancies and minimize development requirements, The PC receives alternating feedback from the in-game environment and from societal interactions after the food choices and at the end of the game.

2025-04-28

2025-05-30

In the following section, we provide brief descriptions of the key outcome measures used in our study.

We measure knowledge about the environmental impact of food systems using a 7-item quiz questionnaire adapted from Hartmann et al. (2021).

Attitudes toward sustainable food consumption are measured in three ways: (a) intentions to switch to a more sustainable diet and lifestyle; (b) support for systematic government intervention promoting sustainable dietary choices as well as support for two concrete policies (food labels and environmental taxes); (c) willingness to forego a bonus payment in favour of donating to an environmental or a sustainable food charity.

Behaviour is measured through an incentive-compatible grocery shopping task conducted in an experimental supermarket featuring approximately 300 food products available at Tesco, the UK’s largest supermarket chain. Based on Clark et al. (2022) methodology, we will compute the environmental impact scores (EIS) for each participant's selected shopping basket. To further incentivize grocery choices, we offer a randomly selected subset of participants the opportunity to receive their chosen shopping baskets for free after the study.

Our main outcome measures aim to capture knowledge, attitudes, and behaviour with regard to sustainable food consumption.

Knowledge: To measure knowledge about the environmental impact of food systems, we developed a 7-item knowledge questionnaire which includes questions on the environmental impacts of different food items. The questions are partially adapted from Hartmann et al. (2021) and modified to also reflect information given to participants in the video game.

Attitudes. We measure attitudes toward sustainable food consumption in three ways. First, we collect self-reported intentions to adopt a more sustainable diet, reduce meat consumption, and make lifestyle changes. Second, we will elicit support for systematic policies that are aimed at facilitating a switch to more sustainable diets and food systems. This includes general support for the government taking action to reduce meat consumption, as well as support for two specific policies, namely introducing environmental impact labels for food and taxing food products based on their environmental impact. Third, we include a donation opportunity to environmental or sustainable food related charities (the Woodland Trust, Rethink Food). Specifically, participants will be asked whether they would like to keep a £1 bonus payment (awarded to participants who complete all three surveys) for themselves, or instead forgo it in order to authorize a donation of £1.33 to one of two charities.

Behavior. To measure food choice behavior, we conduct an incentivized grocery shopping task, using an experimental online supermarket. Previous research has shown that choices made in experimental supermarkets are predictive of real world choices (Howe et al. 2024, Crosetto et al. 2019). Our pilot also suggests that in 9 out of 10 food categories, the shopping basket choices are strongly significantly correlated with self-reported real-life consumption. We use the "Woods'' research platform which was developed by the University of Oxford and mimics a real online supermarket. We populate it with about 300 food products available at Tesco, the largest supermarket chain in the UK. These items were selected from the following categories: fresh fruit and vegetables; fish, meat and plant-based alternatives; eggs, dairy and dairy free alternatives; ready meals. Recent product pictures and prices are extracted from the Tesco website. Subjects receive a virtual budget of £50. We incentivize choices by informing individuals beforehand that 10 randomly selected individuals at the end of the study will have the opportunity to receive their chosen shopping basket for free. Thus, participants have incentives to select products that they actually would like to receive for consumption. The Woods supermarket platform records information on every action on the shopping basket that participants take, such as adding items, removing items, and checking out. To evaluate the sustainability of participants’ grocery choices, we use information from Clark et al. (2022) to assign an environmental impact score (EIS) from 0-100 to each product, where higher values indicate larger negative impacts. Anticipating a right-skewed distribution, we will also apply a logarithmic transformation of the EIS.

In the following, we describe additional post-intervention measures that we include to help shed more light on intermediary outcomes and behavioural mechanisms through which the video games can influence attitudes, knowledge, and behaviour.

Entertainment experience: Immediately after participants complete the video game, we gather their feedback and reactions using a tailored set of assessment tools. We do so using a shortened version of the Modified Differential Emotions Scale (mDES) to gauge emotions such as happiness, optimism, reactance and guilt. Additionally, we administer closed-questions on the game's content and gameplay along with a condensed version of the Narrative Engagement and Transportation Scale.

Educational experience: We assess participants' perceptions of the game's main theme through an open-ended question. Additionally, we ask participants to rate the game's perceived informativeness and potential political bias, as well as whether it encouraged them to reflect on their own real-life behaviors.

In-game choices: Using the game completion code provided by participants, we can track their in-game choices and determine whether they experienced the "good" or "bad" ending. .

Climate concerns: We collect data on the level of concern about more general issues related to climate change and the environment, adapted from Leiserowitz et al. (2019) and Poortinga et al. (2018).

Perceived efficacy: We will assess the extent to which participants feel capable of taking personal action against climate change (individual efficacy), and their belief in society's collective ability to combat climate change (collective efficacy), using measures adapted from Poortinga et al. (2018).

Norm perceptions: Participants will be asked to estimate: (i) the current proportion of the UK population that reports trying or having tried to reduce their meat consumption (descriptive norm), (ii) the proportion that believes people in the UK should reduce their meat consumption (injunctive norm), and (iii) the share of the UK population expected to have reduced their meat consumption in 10 years (dynamic norm). Responses are recorded on a continuous scale from 0 to 100. The wording of these questions follows the Social Marketing Foundation's survey conducted in September 2023, enabling benchmarking for (i) and (ii).

Grocery shopping task motivations: Following the experimental online supermarket task both in the intervention and follow up, participants will respond to an open-ended question asking them to briefly describe their reasoning behind their product choices. This allows us to assess, in an unprompted manner, the impact of the game on participants' attention to the sustainability of food items.

Perceived impact of shopping basket: After completing the shopping task, both in the intervention and follow up survey, participants will be asked to estimate the average environmental impact of their selected basket in terms of decile of the distribution using a slider. Estimates are incentivised through a bonus reward of £0.50 for the correct answers relative to the participants in the pilot study. This captures participants’ beliefs about the sustainability of their product choices.

This study tests whether using interactive play can influence knowledge, individual attitudes and behaviour and which type of educational content is most effective. Our aim is to compare the effectiveness of treatment versions to a control game version while also looking into pairwise comparisons and across differing time spans in a within-subjects survey-based experiment.
The study in fact employs a longitudinal design involving three surveys: a baseline survey, an intervention survey, and a follow-up survey. In the baseline survey, we will collect socio-demographic data and assess participants’ baseline knowledge, attitudes, and information on the food consumption behaviour. In the intervention survey, one week after baseline is completed, participants will be randomly assigned to play one of the four game versions in a between-subjects design. After completing the game, the outcome measures are collected. The follow up survey will be conducted some weeks (plausibly around 2 or 3 weeks) after the intervention. Participants will be asked again to answer questions and complete the necessary task to elicit the outcome measures once more. This allows us to determine whether the effects of the game are short-lived or have some persistence over time.

We will divide the sample into 2 subgroups based on their prior meat consumption and then randomly assign participants within each stratum to different treatment groups using the computer. During the recruitment stage, participants will also be randomly assigned, based on the parliamentary constituency of residence, to receive either a message that highlights the opportunity to try a video game or a neutral message that does not.

stratified randomization at the individual level

No

0

around 3,600 individuals.

We aim to recruit a final sample size of around 3,600 participants who complete the full study (approximately 900 per treatment arm). Since our study has a longitudinal structure and we anticipate around 25% overall attrition we will start recruiting 5100 people in baseline.