Revisiting domain general vs domain specific

Does adopting an extreme domain specific position really reflect the evidence?

Photo by Kaleidico on Unsplash

Summary

• The ongoing debate between domain general and domain specific skills has spread into the educational leadership domain.

• An important plank of the domain specific case is the claim that there is no evidence that domain general skills are trainable.

• This post will give three interesting examples of trainable domain general skills as yet largely unexplored in the context of school leadership.

• There are other issues with the domain specific only position, but interestingly these examples are also a meta-problem for the domain general position.

Establishing context

The teaching of domain general versus domain specific debate has been around for at least a decade and has recently spread to the educational leadership domain. In the context of educational leadership, the debate between the domain general and domain specific positions can be characterized, and hopefully not caricatured like this: On the one hand, the domain general position argues that there are general leadership skills that apply across domains and that therefore can be usefully learned regardless of your profession. On the other, the domain specific position would say that specific domain knowledge is what matters and that even apparently domain general skills such as vision creation are underpinned by substantial domain knowledge.

Rather than revisit this debate, this post will respond to one specific aspect of the domain specific position, namely the challenge to the domain general position to give an example of trainable skills that can have an impact across domains. Here are three in response.

Goal setting

Considering how ubiquitous goal-setting is to school leadership practice, playing a role in everything from school improvement, to target setting, to performance management, goal setting research has been oddly under explored by the school leadership literature. Goal setting is both a trainable skill and more importantly for the current discussion there is some increasing evidence that goal-setting has a domain general “spread” effect. For example, writing life goals can lead to improved GPA, and more generally writing life goals leads to improved academic performance.

Superforecasting

@PTetlock’s work on expert political judgement and the Good Judgement Project suggest that domain experts can be outperformed by domain generalists in the field of forecasting, and that forecasting skills can be improved by even short training sessions. This is particularly potentially relevant for school leadership where training in strategic thinking is often lacking, and as most strategic decision making in schools, particularly with regards to the external operating environment of the school, is essentially a matter of best bets forecasting based on the available evidence. With specific techniques such as “Fermizing” questions, balancing inside and outside views and carefully monitoring track records, forecasting training has a great deal of potential to offer school leadership research.

Perhaps most importantly, prediction tournaments (alongside small scale trials) offer a tantalizing glimpse of a process for making and implementing better educational policy globally. The wonderful Metaculus site has yet to adopt education as a topic beyond some generally US higher education questions so this is an area ripe for further development.

Cognitive debiasing

The third of our three interesting areas is that of cognitive debiasing. Based on the idea that there are some inherent limitations in human reasoning that would therefore apply across domains, cognitive debiasing to both raise awareness of these and suggest strategies to counter them would be another example of a trainable domain general effect that could be an exciting area to explore for school leadership research. Most pertinently for the discussion, there is increasing evidence for the efficacy of cognitive debiasing training including laboratory studies, and more importantly at least one recent field study.

Conceptual issues

Summing up, despite the target of the above there are most likely conceptual problems with both positions, especially arising from the links to evolutionary psychology made in the original Tricot and Sweller article which has resulted in some very interesting implications they probably did not intend. In particular, the role of intra-specific competition in human evolution implies that humans have evolved to collaborate and compete, and as those evolved behaviors are very likely to impede effective group collaboration in the modern workplace then it logically follows that collaboration should be taught precisely because we have evolved to do it rather than the reverse. This is a topic deserving of its own post, but some examples of those behaviors include: we are more worried about signaling group membership than being competent, tend to defer to the opinion of the highest status person in the group regardless of the quality of their contribution, combine to attack high performers in our groups who reduce performance in response, self-deceive in order to better deceive others, and that we will even incur costs to ensure others in our group get less rewards than ourselves.

Lastly, while this blog post may appear to be more critical of the extreme domain specific position, taken as a whole it also highlights a meta-problem with the domain general position as the above points are typically not included in its articulation. This leads to a final summary point, that regardless of where you fall in the domain general versus domain specific argument in the context of school leadership, there are a great many interesting research questions to explore.

Further reading

Goal setting

• @bennewmark ‘s classic blog explains why outcome based target setting in schools may not be the best idea. The fact that this literature based critique came from a practitioner blog rather than education research supports further the idea that this is a potentially very fertile area for school leadership researchers to explore.

Superforecasting

• Join the good judgement open or Metaculus and learn by doing or read Phil Tetlock’s book.

Cognitive debiasing

• There is a very accessible summary of the field study here.