Responsible Modelling

Responsibility is fairly difficult to define precisely, if we see it as complementary to but distinct from legal commitments and other formal accountabilities. Professional responsibility has that character, a behaviour which is not codified but recognisable as a signifier of competence beyond technical skills. And one of the tacit societal expectations of professions is that they self regulate the standards which are specialised to the profession to assure benefits to society. In general, professionals need to keep up with the development of new practice, and contribute to it, assuring it meets the needs of society.

Modelling has a contentious role in public policy, not just in public health but also climate and economics, particularly in respect of future projections. Computational models have substantial value but relying on them puts a great burden of responsibility on the professional process developing them to specify, derive and communicate them well. So professionals ought to assure the public that modelling is used responsibly, in order for them to enjoy legitimacy in public policy applications. And they need to be engaged and responsive with how they are used, and their impacts on the public, to guide improvement.

Models are used as a structured way to things we do not know and cannot know, for scientific, temporal or empirical limitations. Concerns that we could be better off with the true values were made about the public accounts when parliament first started to ask for estimates in the 17th century. But at that time, expert calculations were publicly promoted and pored over, for their reputation as well as their value. And this reliance on professional expertise remains a feature even if it rarely comes to popular attention, but it warrants more consideration.

A responsible modeller ought to be honest, competent and reliable, by the standards of fellow professionals, and thence trustworthy for the public. But the specific knowledge role of models has a substantial determination of how risk and opportunity is ununderstood in substantial areas of public policy. So there is a political accountability in models, going beyond their robustness or the expertise of professionals developing them. This is not just for the accurate and consistent use of a model but the choices that went into commissioning it, and not alternative specifications.

Communication is at the heart of the challenge, between experts, politicians and publics, often mediated by others and not actually about technicalities. But this requires acknowledgement of what can be communicated, say provenance, assumptions, values, uncertainty and implications, and the consequences when this is disrupted. And each one of these is difficult to ascertain, specify and describe, which often leads modellers to obfuscate, oversimplify or overlook audiences entirely. Unmediated, structured stakeholder interaction is hard, and unfamiliar to those in very technical practice.

Mediators are less likely to be honest brokers than bringing partial interpretations in the sense of preferences as well as completeness. But an aim for concomitant legitimacy will rely on including people at some stage, so excluding them otherwise is a risk. So democracy can get lost as the technocrats take on responsibility for each other rather than bringing it back to the public. But values, and how these are represented in assumptions, are not technocratic issues, and this limitation can be exacerbated by narrow attitudes of some mediators, who really believe science is objective.