Gamified agricultural learning: A randomized controlled trial on cowpea knowledge and intercropping adoption.

The intervention begins with the implementation of the educative board game for the treatment groups and oral information on cowpea and intercropping benefits for the control groups. After the intervention, all four groups will answer knowledge questions regarding the benefits of cowpeas and intercropping, focusing specifically on soil health, nutrition, and climate resilience. Finally, cowpea seeds are distributed with advice on best-management practices to all participating farmers to facilitate practical field implementation.
The treatment: The game is played with an enumerator who guides the respondent through five agricultural years. The board represents the farmer's field and household, and local beans are used as game pieces. Next to the board is a visual reference table showing the costs, harvest amount in a normal year, harvest amount in a year with a dry spell and effect on either soil or diet of three crop choices: Cassava alone, Cowpeas alone, or Cassava-Cowpea intercropping.
To learn the rules, the farmer is guided through a mandatory first year of planting cassava alone. The farmer takes game pieces from their "storage" to pay for the planting material. After the crops have grown, they can be harvested and the points will be placed in the “harvest field”. The farmer must immediately move 4 game pieces from the “harvest field” to the "family field" to feed his family for the year. The remaining game pieces go into storage.
From the second year on, the farmer gets to make their own choices. At the start of each year, they go to the "market" and decide what to buy based on their budget in storage. They can stick with cassava, or choose cowpeas, or cassava-cowpea intercropping. The farmers will learn about the consequences of their choices for the soil health or the dietary health. The treatment group which plays the soil focused game will be faced with the tradeoff that cassava planting might bring a lot of yields but reduce the soil quality. Respondents in the nutrition focused game will learn that eating only cassava will not suffice for a balanced diet. Both games actively promote the cultivation of cowpeas either mono- or intercropped by improving the respective scales. All game mechanism are exactly the same for both groups, only the storyline behind them differs. The enumerator updates the board based on the choice, moving game pieces around for costs and harvests, and adjusting the soil health or diet scale. In year 4 a dry spell will hit the farmers. The farmer faces reduced yields, but they can experience how cowpeas or intercropping survive the drought much better than monocropped cassava. At the end of year 5, the enumerator counts all the remaining game pieces in storage out loud. Finally, the farmer is shown a "Scenario Board," with best-case and worst-case scenarios to compare their own performance. Crucially, the farmer sees that planting only cassava in every year is not sustainable since their soil health points or their family diet points, depending on which version of the game they are playing, decrease. After the game is finished, the farmer receives their performance-based in-kind pay-off.

2026-06-29

2026-07-31

• Knowledge: Knowledge outcomes are measured using a standardized index constructed from a combination of true/false questions regarding the agronomic, nutritional, and climate-mitigation benefits of cowpeas. We evaluate these cognitive outcomes regarding short-term and long-term effects.
• Cowpea adoption: A binary indicator equal to one if a farmer cultivated cowpeas by the endline survey, and zero otherwise. Additionally, the total land area allocated to cowpea cultivation to compare the intensity of cultivation between the treatment and control groups.
• Intercropping adoption: Intercropping adoption is defined as the cultivation of cowpeas mixed on the same plot with any other crop, meaning it includes but is not limited strictly to cassava-cowpea systems. This outcome is analyzed as a binary indicator at endline. To isolate the behavioral shift driven by the information delivery methods, we will evaluate these endline intercropping choices against the baseline intercropping habits recorded for each household. We will also evaluate changes of the land surface cultivated using cowpea-intercropping to measure changes in intensity of adoption.
• Cowpea and cowpea-intercropping intentions: This variable measures the intentions of farmers regarding the future adoption or continuation of cowpea intercropping. It is captured right after the intervention as well as at endline using a structured metric such as a 5-point Likert scale assessing the respondent's stated willingness to practice cowpea cultivation and intercropping in upcoming agricultural cycles.

• Change in how often a person worries about their soil
• Change in cowpea perception
• Change in consumption of pulses (amount and frequency)

This study evaluates how different topical foci within an educative game affect knowledge retention and adoption behavior. In addition, the study assesses the broader feasibility and suitability of using the educative board game as an information delivery method. We deploy a serious board game to familiarize farmers with the value of cowpea cultivation and cowpea-cassava intercropping. We use a 2X2 design, where we vary both the content focus of the information provided (one emphasizing soil health benefits and the other focusing on nutritional benefits) and the mode of delivery (short oral information treatment vs. serious board game). Both game versions actively promote cassava-cowpea intercropping as a sustainable practice to enhance land productivity and mitigate climate risks. We compare the gamified treatments with the control group receiving oral information of cowpea and cowpea intercropping benefits, and the group receiving the nutrition framing with the group receiving the soil health framing. Within the intervention, game participants are incentivized to maximize their game outcome through performance-based, in-kind payoffs. Meanwhile, control group participants receive an average in-kind payoff to ensure equity. After the intervention, cowpea seeds will be distributed to all participants. By pairing an immediate post-intervention assessment with a subsequent endline survey (administered via phone), we capture short-term knowledge gains, long-term retention, and both intended and actual adoption. This design allows us to isolate how the specific topical foci and the delivery format drive variations in knowledge retention and sustainable practice adoption.
We organize our analysis around three overarching research questions.
1) How does the framing of information regarding cowpea cultivation and intercropping influence farmers' short- and long-term knowledge, adoption intentions, and actual adoption behaviors for climate adaptation?
We hypothesize that the informational framing will differentially affect knowledge retention and adoption behavior. Building on the agricultural technology adoption literature, we argue that information provision alone is not a sufficient predictor of adoption. Rather, the cognitive processing of innovation depends heavily on information framing and salience effects (Piñeiro et al., 2020; Rust et al., 2021).
As a sub-research question we additionally analyze whether the framing of information within the game matters: is the nutrition game or the soil health game more effective in influencing farmers’ short- and long-term outcomes?
2) How does the effectiveness of the two information framing treatments differ based on farmers' socio-economic characteristics, including gender, perceived soil quality and household food diversity?
Information highlighting nutritional value is expected to resonate most strongly in contexts of food insecurity and among female farmers, who are traditionally responsible for food preparation and allocation. Conversely, information focusing on soil health is hypothesized to appeal to farmers managing degraded plots and to differ across different agro-environmental zones.
3) What are the comparative impacts of playing a serious board game versus receiving standard oral information on farmers' knowledge, adoption intentions, and actual adoption behaviors?
Here, we investigate mechanisms of active versus passive learning in agricultural information delivery. Traditional extension methods rely heavily on oral transmission, which can suffer from high cognitive load and rapid decay in retention. We hypothesize that the educative board game will significantly increase both short-term recall and long-term knowledge retention compared oral information.
We extend our analysis from purely cognitive outcomes to adoption. We hypothesize that the gamified intervention will increase both the adoption rate of the distributed cowpea seeds and the propensity to practice intercropping. Our hypotheses are based on evidence that shows that how an agricultural practice is communicated alters its perceived relevance and farm-level utility (Carrico et al., 2015). We define cowpea adoption as the actual cultivation of the distributed seeds. To evaluate this outcome thoroughly, we will analyze both the binary decision to plant and the total land area allocated to the crop across the treatment and control groups. Furthermore, our definition of intercropping adoption is not restricted exclusively to cassava-cowpea systems; it encompasses any intercropping combination that includes cowpeas. To capture this behavioral shift, the empirical analysis will directly compare intercropping practices from baseline to endline. The mechanism driving this hypothesis is experiential simulation: the board game allows farmers to test farming decisions, observe hypothetical crop yields under normal and under drought conditions, and experience the risk-mitigating effects of intercropping in a low-stakes environment. This simulation is expected to lower the perceived risk of intercropping techniques. When paired with the subsequent seed distribution, farmers who played the game are hypothesized to transition from information acquisition to practical field implementation at a significantly higher rate than those in the oral information control group. Our study hypothesizes that participation in the educative board game will result in higher adoption intentions of cowpea cultivation and cowpea intercropping in the future. This will be measured directly after the game as well as at endline. Measuring adoption intentions at endline is methodologically important because actual field adoption during the immediate study timeframe may be restricted by temporary household labor, land, or seasonal constraints.

We anticipate distinct heterogeneous treatment effects across two dimensions: respondent gender and agro-environmental conditions of the respondent’s location
1. Gender
• Anticipated Effect: Female farmers will display higher knowledge retention, stronger adoption intentions, and higher actual field adoption when exposed to the nutrition game, while male farmers will respond more strongly to the soil game.
• Rationale: In rural Madagascar, intra-household labor, decision-making, and crop management are highly gendered. Root crops like cassava are traditionally viewed as men's crops, whereas legumes like cowpeas are more frequently managed and harvested by women. Because our study promotes intercropping cowpeas directly within existing cassava fields, the proposed practices intersect directly with these gendered divisions of agricultural land. Furthermore, women bear primary responsibility for household nutrition, child health, and dietary diversity, making the framing of the nutrition game highly salient to them. Conversely, men focus heavily on primary plot management and long-term land productivity, aligning closely with the soil-health messaging.
2. Agro-environmental conditions
• Anticipated Effect: We expect differential effects depening on the agro-environmental and climatic conditions of the farming households. We hypothesize that farmers in more arid areas (likely in regions Androy and Anosy and south of Atsimo Atsinanana) will display higher adoption intentions, and higher actual adoption. Conversely, farmers in more humid areas (Atsimo Atsinanana) will display the largest relative learning effect.
• Rationale: According to our data, cowpea production is very common in semi-arid Androy and Anosy, but only practiced by a minority in humid Atsimo Atsinanana.
We also expect treatment effects to vary with prior cowpea cultivation experience, household characteristics (age, education), perceptions of cowpeas and intercropping, social capital and group memberships, access to extension, perception of soil health, food diversity and risk preference. We explore heterogeneity along these dimensions by interacting the treatment with each of these characteristics (one-by-one).

Not available

To ensure internal validity, treatment status will be assigned ex-ante using a reproducible Stata script. Randomization is done at individual level, the treatment conditions will be administered fully in private and all respondents receive cowpea seeds, so we expected very limited spillover risk, if any. The randomization will be stratified by fokontany and the gender of the respondents, allocating eligible male and female household members with equal probability to either the nutrition game, the soil health game, or the control group, and subsequently with equal probability to one of the control group conditions. These assignments will be pre-loaded directly into the survey software prior to data collection. When enumerators arrive at a household, their tablets will automatically pull the household ID data and display instructions specifying exactly which member to interview and which intervention to administer. In case of assignment of the intervention to the second respondent but they are absent during data collection, the participation of the experiment will default to the main respondent.

individual

No

63 fokontany (smallest administrative unit)

1038 respondents

346 soil game (treatment group 1), 346 nutrition game (treatment group 2), 173 oral information with soil health focus (control group 1), 173 nutrition information with soil health focus (control group 1)

When evaluating main treatment effect of comparing the soil intervention groups (n1=346) directly against the nutrition intervention group (n2=346), the study is fully powered to detect an MDE of 0.21 standard deviations. To compare the serious game intervention groups (n1=692) against the oral information control (n2=346), the study is powered to detect an MDE of 0.18 standard deviations.