Raising Concerns

Statistical data ethics is an applied ethics, of what ought and ought not to be done, and as such eventualities as ought not to be done potentially transpire. While some cases have suitable oversight and therefore only needs support in terms of insisting ethical requirements are acted on appropriately, others lack such oversight for several different novelties. These can relate to unprecedented circumstances, as presented in the early stages of the SARS2 pandemic, and extreme urgency which may have been present in the pandemic and may have seen reckless behaviour. So existing oversight may not be equipped or agile enough to respond to challenges, but the transformative utility of data also brings new ethics.

Transformative Applications

Broadly, data is transformative for three separate reasons, all of which mean governance structures may not be adaptable to the new applications, or in some cases even absent.

First, data is being repurposed, whereas statistical uses were based on designed primary data collection, transformative applications use data which arose in other administrative processes or delivery systems, which has error structures due to its origins, and sometimes complex classifications and exclusions.

Secondly, the scale of data used in transformative applications, which can allow for greater adjustments for the biases in the first point, can put substantial restrictions on direct scrutiny of individual observations which are typically advised in statistical training, meaning professional have to adjust their practice, including developing new tacit knowledge of how infelicities manifest.

Lastly, transformative applications can be radically transformative of the sociotechnical systems they analyse such that distortions are caused at a higher order than familiar ideas of satisficing in survey responses and performativity to targets, and these patterns will be unfamiliar to those non-data professionals implementing changes.

So it may be that transformative applications of data have unintended and even disastrous consequences that are not easy to interrogate, or even readily apparent, for administrators. As such it is incumbent on responsible professionals to raise concern, from whichever vantage point they perceive a problem, but also necessary to establish that it is escalated to a level where its significance is appreciated. As this appreciation extends to the nature of the impact and the role of the transformative application in contributing to it, both of which may be complex and unfamiliar, this means quite a lot more is involved than saying there is some kind of problem, and some useful steps are described below.

Issue, scale and focus

Essentially there are three steps which isolate what to raise a concern about, which is difficult to do at first glance, typically because of the transformative nature of the issue:

• Whatting the what – words which do not already set a frame are elusive, but the point of this is to establish the ontology of the issue and as necessary the sociotechnical transformation causing a concern, and understanding this means contrasting precursors, expectations and practice.

• How big is the picture – the transformation can impact on a system vastly greater in sociotechnical scale than the analytical team responsible for doing the development work, and as this is a transformation it can be seen to set a precedent outside the specific application.

• The heart of the matter – although a concern may arise in an analytical setting, the size of the picture may be larger, to the part that the principal interest stands elsewhere, further a transformative application has many facets, so the root cause is what signifies.

Essentially this means when something is happening which ought not to be, identifying what it is, how it is happening and the key issue which is at stake. This leads to a sense of proportion, easy to misjudge in transformative applications (in both directions), and a recognition that aggravation often perceived in interrogating the issue which is likely not to be an aggravating factor in its significance. Raising a concern then means mentioning this in such a way that it is appreciated and resolved, which is very involved – if it weren’t it would be done already.

Raising the concern

The engineering profession recognises a professional duty to the public interest, and guidance describes this as a duty to warn when professional judgement is compromised in some way. Although this is not available at present, the decision is presented as questions of whether, when, who and how to warn. Having decided, the question of when may be more about escalation rather than biding time, but it also indicates that gaining more information is prudent. In respect of who, the scale and focus are salient, so it needs to be raised with some body with ability to appreciate the issue and responsibility to act.

When the body already has ability and responsibility, drawing a concern to their attention implies its significance had not been appreciated at that time, and that could indicate a degree of negligence. But given the transformative application of data, it will also mean it is difficult to perceive, and therefore a clear explanation is important, as well as a concern that is about salience and significance. So being direct and unambiguous is important, and indeed writing this out for someone who can recognise this.

Avoiding misinterpretations and defensive reactions is critical, and so setting out what is not the problem can be helpful, as well as how this is distinctive, and readiness to further explain as opposed to pressing the significance. The possibility of a negative reaction for implications of responsibility to negligence or business and reputational stakes is important, not only for self-preservation, but also as it is a distraction and much easier to attack the critic than engage with the issue which has prompted the concern.

Escalation of concerns

Raising a concern is about transferring it to a person who will take action, even if they are acting on behalf of a body, and they may determine not to do so. Where this determination does not make sense, ethics dictates the concern should be escalated to be taken seriously, and if the representative of a body feels nothing is necessary, it does not make sense to ask the same organisation. Particularly, it is best to avoid raising the same concern in different ways with the same person, akin to the futility of repeatedly banging one’s head against the same wall.

As ethics is unfamiliar, difficult and acute, with not acting meaning maintaining business, a response that no action is necessary is to be anticipated, and further escalation must be planned for. However, where a response gives more information about factors considered, it is right to take these into account, and also direct the concern to others who will appreciate what remains at stake. Although a first approach ought to engage the body initiating the problem, further escalation needs to be pursued independently of their interests.

Although regulators and ombudsmen have responsibilities to handle complaints independently, they will have a process set up for known types of issues, not the impact of transformative applications of data. So drawing a concern to their attention needs to fit it within their own process, and also strategic considerations where the concern is partly about the precedent set. But the key to this step in escalation is there is a substantial independence in considering the issue beyond proximal business pressures.

Resolution

If a resolution is emerging, it can still be difficult to put into practice, particularly taking time. So a firm commitment (formal but not necessarily a written response or public challenge) is required so that it is followed through. However, as the issue was about the ethics not more typical issues of regulatory enforcement, the delivery is out of scope of a concern. This may be frustrating, but as the problem was about the issue, not the wilful malevolence of individuals, it is not entailed to pursue this final step aggressively.

Of course ultimate authority over any matter can exist, and determine against the concern being resolved along the lines of an ethical analysis. Holding on to a denial of this kind will be acrimonious at least, so a final step is to refer the matter to a moral authority, whose view all concerned would accept, even as there is no actual supervisory role. Identifying who to escalate and refer matters to is perhaps the hardest part of raising concerns, but gives a fascinating insight into the upper echelons of the administrative state.