A holistic approach to improving development decisions

Research in agriculture seeks to generate new approaches or technologies that can be used to make a difference in farmers’ lives, and for the broader society. Well-designed technologies can have major positive impacts, as well as providing evidence that investments in agricultural technology development pay off.

But translating research-derived technologies into impact on the ground can be a challenge. Especially when technologies come from on-station trials or small pilot areas, it is often unclear how they will perform under ‘real-world’ farm conditions.

Forecasting the performance of a technology requires consideration of site-specific conditions at the target area and in the target population, which may differ substantially from those at the place/s where the technology has been tested. But since it is impossible to test-drive innovations in all environments, we have to find alternative ways to forecast their performance.

A good first step in this direction is extensive interaction with local informants and experts on local systems, whose insights can be invaluable for anticipating potential adoption problems. In this interaction, it is important to seriously consider all factors that are brought up, even when these are unexpected or difficult to quantify. This is because technology adoption and performance can be limited for a host of reasons, including biophysical, socioeconomic, and cultural factors.

The adoption process is critical for dissemination of new technologies, and careful planning for impact is needed. Many adoption problems can be avoided by asking early in the technology development process how exactly adoption will happen and what can be done to strengthen it.

For a greater chance of success we must carefully plan for impact. As we start working on a new idea, we can already begin to work out the uncertainties and adoption bottlenecks and take action to prepare the ground for the new technology before it is ready for dissemination. Based on what we find, it might even be necessary to adapt the technology based on realities on the ground.

Expressing our impact pathways

As researchers, we also need to get better at articulating how we achieve impact. Developing a quantitative impact projection model at the beginning of a technology intervention is a good way to do this. Such a model can help plan for impact and manage risks by anticipating problems and taking steps to mitigate or avoid them.

The model should consider all the factors that are important for the success of a technology, and it should include all risks to this success. Who needs to be on board? Whose attitude must change? What biophysical environment is required? What markets are needed? Do cultural norms have to change?

Such a model should also aim to account for our uncertainties. For instance, how will future weather affect our technology?

It also has uses for monitoring and evaluation: we can update a model as reality unfolds. Say we face adverse weather in a given year. We can tell this to our model, and it will automatically adjust our impact targets. This allows us to account for effects that we have no influence on, such as weather, political events or other factors, which can make our technology look great or terrible for no fault of our own.

Decision analysis can support us in keeping science relevant

If we want impact, we have to influence decision-makers. The most influential research is research that supports good decisions.

What information is needed for particular decisions is very variable and difficult to foresee. It requires a thorough analysis that explores all the things the decision-maker needs to factor into a decision, what is already known about these things, and how strongly they affect projected decision outcomes.

At the World Agroforestry Centre (ICRAF), we use decision analysis to make such models. Decision analysis is an approach that is optimized for supporting risky decisions on complex systems under uncertainty. It is a tool widely used in the business world and sometimes in policy analysis, but one that is only now entering agricultural development research.

To provide tailored support to specific development decisions, we develop causal decision models to simulate the likely impacts of alternative decision options. These models are developed based on various sources of information, but drawing heavily from the inputs of experts, who are convened in a model development workshop. Facilitated by decision analysts, they bring to the table everything they think decision-makers should consider, and assemble these factors into a causal model. Whenever possible, the decision-maker should be a part of this model development group.

Once a model exists, the current state of knowledge about all the input variables is assessed. Where no precise numbers exist, which is often the case, values are expressed as confidence intervals or probability distributions. With such inputs, we can run probabilistic simulations, which can convert uncertain inputs into uncertain outputs. This means that they give us plausible ranges of expected outcomes. Sometimes such ranges offer enough guidance already for which decision option holds the greatest promise for impact.

Quite often, however, we find that both positive and negative outcomes are possible, when we base our assessment on what we currently know. In such situations, we use ‘Value of Information analysis’ to determine which knowledge gaps we should do research on. Normally, there are a few variables that carry most of the information value, identifying them as decision-specific research priorities.

Holistic system valuation through Decision Analysis

Decision analysis also has a role in ensuring environmental sustainability is taken into account as we increase food production. We have to get better at considering the value of all environmental services in our decision-making. Crop yields, surely, cannot be all that matters.

But considering things like hydrologic services, biodiversity conservation or even the cultural or spiritual value of a farm is difficult, because these cannot be as readily measured as the crop that is harvested. That is why all too often, we let the lack of a precise number lead us to not counting these services at all. But that is like saying their value is zero, which is obviously not the case.

To tackle this issue we need to embrace working with ranges rather than precise numbers for everything. We can, for instance, come up with plausible ranges of values for ecosystem services that accommodate different perspectives or different people’s estimates. These ranges can then be used for cost-benefit calculations.

We then, of course, do not get precise numbers as the result of such calculations, but also ranges, which again may include high and low values. We still wouldn’t know which precise number to pick, but the ranges (which are really probability distributions) would allow us to compare the relative benefits of, say, different land use options in a holistic way – a way that considers all the environmental services.

By taking the system as a whole, we will probably come to realize that multifunctional agriculture with trees, from which farmers gain benefits such as fuelwood, shade and timber, offers more value than a monoculture, even though the latter might have higher crop yields.

But we can only come to this conclusion if we adopt a holistic view on systems that considers all the services they provide. Through its capability to consider ecosystem services we cannot quantify precisely, decision analysis can greatly support the causes of productivity, multifunctionality, and sustainability.