Some methods related to causal mapping and even some forms of causal mapping itself are not interested in individuals' different causal perspectives: they are primarily aimed at reaching a consensus, expert map straight off without first recording individual viewpoints and attempting to combine them.

You have a causal map with lots of links from an intervention to a final outcome. It's a really impressive chain.

We have often processed data gathered specifically for evaluation purposes using "causal back-chaining". But we often process secondary data which was not specifically intended for causal coding.

When gathering primary data, the way in which questions are asked influences the meaning of the maps and their links. For example, in the QuIP, (Copestake et al., 2019b) respondents are asked to identify causes of changes, then causes of the causes and so on. This means that most of the factors are already as changes in something, such as ‘an improved harvest’ or ‘reduced hunger’). This has implications for how positive and negative statements are combined, as discussed later.

In a QDA study which interviews 12 farmers about their own experiences, we can code information from source S like “the new seeds brought me better crops” as information about source S. But what do we do with information like “I think most of us had the same successes” or “the farms in the other side of the valley did not have the same success”, or “the seeds were very productive for me but only in the shadier fields”? These challenges arise especially in causal mapping and are very real in our daily work at Causal Map and Bath SDR.