EvalC3 can be used to explore a number of different types of causal relationships. At the very macro level these fall into these two categories:
- Causes of an Effect: An effect can have multiple causes.
- E.g. There may be a number of factors, which led people to attend a conference
- Effects of Cause: A cause can have multiple effects.
- E.g. Attending a conference may have many different effects on what I do afterward
More often people are trying to identify cases of an effect. These can take various forms:
- Conjunctural causes: Many events are caused by combinations of factors, rather than single factors.
- E.g. I went to a conference in June because I was interested in the subject of the conference, I had friends going there who I would like to see and I had the time available to go.
- Equifinal causes: Events can arise as a result of many different conjunctions of factors.
- E.g. Some other people went to the same conference in June because their boss told them to go and they had the relevant subject knowledge within their organisation.
- Multifinal causes: Particular factors (or combinations of these) can lead to many different effects.
- E.g. People attending the same evaluation conference session end up making use of the session contents in many different ways
- Asymmetric causes: The causes of absent events may not be simply the absence of factors that cause them, but the occurrence of other additional factors.
- E.g. One friend of mine did not go to the conference even though he had the time and was motivated to go. Unfortunately, his child was sick and needed to be taken for a medical checkup.
- Necessary but insufficient causes.
- E.g. Having the relevant expertise to attend a conference may be necessary but insufficient. Permission from one’s boss is also needed
- Sufficient but unnecessary causes:
- E.g. Some people went to the conference because they were invited as speakers
- Necessary and sufficient causes
- E.g. For some people, the combination of being told to go to the conference by their boss, and having the relevant expertise was both necessary and sufficient.
- Neither necessary or sufficient causes:
- E.g. Being bored with what I was doing was not necessary or sufficient to lead me to go to the conference in June. but it may have been influential.
- PS: These can be as important and useful as necessary or sufficient conditions. See this blog posting, specifically the section about satisficing versus optimising
- INUS causes: Insufficient but necessary parts of a configuration that is sufficient but not necessary.
- For example, having the most relevant subject knowledge, among all the others in an organisation may be an insufficient but necessary factor that led to someone being sent to attend a conference. (A+B) or (C+D) leads to E
- SUIN causes: Sufficient but Unnecessary part of a configuration that is Insufficient but Necessary. (A or B) + (C or D) leads to E. I can’t think of an example here 🙂
- Exclusive Or causes: Using a different example from the above, both credit and grant assistance may be sufficient to improve people’s livelihoods. But providing them with both together may be counterproductive. (A + notB) or (notA + B) leads to E
What EvalC3 can do
- Find attributes which are Sufficient and/or Necessary for an outcome
- Find combinations of attributes (aka configurations) that are Sufficient and/or Necessary
- Enable manual tweaking of predictive models to identify the extent to which they are INUS conditions, i.e. if the model fails to perform if they are removed.
- Develop separate predictive models for either the presence and absence of an outcome
- Developed predictive models for the causes of an effect.
- Develop multiple predictive models, each which predicts some but not all of the outcomes in a data set.