Nancy Cartwright began the second installment of her Rotman Lecture (‘Wiser use of Social Science’) with a discussion of the tragic death of 17 month old Peter Connelly, a case of child neglect and abuse that led to a widespread media controversy and to a public inquiry into Britain’s child services and social services programmes.

According to the BBC, the first instance of abuse was reported in December of 2006 and the pattern continued—ultimately resulting in the child’s death in August of 2007. During this time, Peter and his family were routinely visited by social workers, once resulting in the removal of Peter from his mother’s care, only to be returned afterwards. Accusations were made by a former social worker, Nevres Kemal, in February of 2007, that their former colleagues were failing to adequately protect Peter. Following Peter’s death and the subsequent media firestorm, a few child welfare workers tendered their resignations and Sharon Shoesmith lost her job as the local director of Children’s Services.

As is the case after any public tragedy, the bereaved and the public began to search for answers. But how are these answers to be found? And how can we know that we have correctly identified the particular event, or series of events, that caused this tragedy? In “Wiser Use of Social Science: Wiser Wishes, Wiser Policies,” Cartwright not only attempts to answer these questions, but also to uncovered the hidden complexity that lies just under the surface of our most pressing public policy decisions. But, before offering a commentary, I will first give the reader a rough overview of her arguments.

After introducing the audience to the “baby Peter” case, Cartwright explained that there were (at least) two different possible responses. The first kind of response she discussed came from the ministry of education and resulted in the firing of Sharon Shoesmith. As Cartwright explained, the “scapegoating” (her words) of Shoesmith and the disproportionate amount of blame that her actions accrued was an attempt to locate a single cause for the death of Peter and, as a result, to localize our blame and condemnation accordingly. I will call this the “model of individual blame,” as it is marked by a singular focus on one aspect of the series of events that lead to the death of baby Peter. In general, this approach is motivated by a desire to act quickly, and in the process of so doing, conditions a poorly researched response. As a result, this approach means to identify a connection between human error and the perceived need to act quickly to solve a systematic problem. Cartwright also discussed another policy recommendation, which I will call the “early intervention approach.” This approach misjudges the quality of evidence that can be gleaned from the causal nexus. It not only results in another oversimplified account of the conditions that lead to a particular event, but it also appeals to such misinformation in order to license a top-down, and ultimately paternalistic response. I will address each in turn.

Blame in this case was applied at a highly individual level; but given the the size of this bureaucracy it is not immediately clear that an individual could be solely responsible for not intervening before Peter’s death. Such an approach ignores the complex web of social and systemic factors that preceded the death of Peter, and which may have been relevant, contributing factors. Cartwright went on to stress that when blame is attributed in a hasty manner, it may put us at an “epistemic disadvantage” for now we are no longer looking for other factors that might have caused the social workers to remove Peter from his mother’s care sooner. These overlooked causes could be, as Cartwright explained, the high turnover rates for social workers or the large case volume for each social worker. In addition to the epistemic worries, the “hasty” condemnation of a single individual will result in lasting effects on his or her personal life. It ostracized Shoesmith, destroyed her career, and may have caused long lasting psychological damage.

The second response Cartwright identified,  the “early intervention approach,” was exemplified by the response of former Prime Minister Tony Blair. Blair suggested that it should be the policy of child protection services to “identify at-risk families,” and to intervene early. Notice that this policy suggestion tacitly assumes that such predictions can be made with a high degree of accuracy (an assumption that Cartwright will go on to challenge directly). In addition to the epistemological problem of accurate predictions, this approach also has a problematic moral component. By labeling families as “at-risk,” we might actually increase the likelihood of abuse, and almost certainly we will add stress to what is likely to an already difficult situation. Further there are questions regarding the rights of parents, their autonomy, and of systemic paternalism that lurk just under the surface of Blair’s comments. And even if his policy suggestions were effective, Cartwright urges the listener to ask if such policies are indeed morally acceptable.

Even though neither of these approaches are desirable, it is clear that each of these approaches are politically expedient, and that they are popular with both politicians and the general public. Responses to tragic events—especially when a government institution is involved—are often characterized by both quick blame and by suggestions that seem to undermine individual autonomy and to increase state paternalism. But, when one is considering how to create better policies, we need to understand these failings at a higher level of generality. This is exactly what Cartwright did in the remainder of her talk.

She began this part of her talk by posing the question: “What are the philosophical stances that prop up [the model of individual blame and the early intervention approach]?” As I mentioned earlier, the analysis of public policy requires prediction (“What will they produce?”) and evaluation (“What did they produce?”). As Cartwright explained, the two approaches discussed above assume objectivity—that there is a single, correct way to understand how and why things went wrong. From this they then reason that it is possible to be certain about what happened, and indeed this approach “banks on certainty.” This approach results in the illicit belief in a single, narrow causal story.

What’s particularly troubling about this series of assumptions is that the complex causal nexus that, in reality, preceded the death of Peter, gets projected onto a single linear causal story. And it is this kind of projection that isolates errors from the contexts in which they occurred, perhaps ensuring their recurrence. The series of events has thus been oversimplified and is often done at the great personal expense of some of the people involved and, likely to the detriment of many at-risk children who need to us to have a greater appreciation of the complexity of situations like these so that we can better help them in the future. If the death of Peter was the result of a complex network of causes all obtaining at the wrong place and the wrong time (which is much more likely than it merely being the fault of an administrator), it behooves us to try to adapt our policies accordingly, and to correct our methods of policy evaluation.

In particular, the problem inherent in the linear causal model overlooks the “helping and supporting factors.” These are the factors that had to be in place for the situation to obtain as it did. Each of the series of causes that lead to Peter’s death required a number of preconditions to be met, and all of the ingredients had to be in place for the outcome to obtain. Given this fact about how events occur, we need a systematic method of analysis when evaluating public policy (to be discussed later). The systematic method can be, at times, unnerving. Regarding social policies, a major implication of the systems approach is that principles are not always easily transferable. What works well in one area may prove to be a resounding failure when the program is expanded into a new community. Because transference is not guaranteed our ability to predict how well a particular policy will work based on past experience is deeply under-determined.

This level of underdetermination is not a failing of Cartwright’s view. It is, instead, an acknowledgement of the precarious epistemological situation we have always found ourselves in regarding the generation, evaluation, and implementation of public policy. Cartwright’s analysis of the philosophical assumptions inherent in the responses to Peter’s death show a clear need for a more careful, detailed approach. And this strikes me as the correct response. After we dispel the myth that answers come easily, that there is only ever one person responsible, that there was an objective series of events which necessarily brought about a particular tragic result, then we are better positioned to get a sense of the myriad causes at play—but also the systemic problems that failed to identify those causes as they arose.

As I said before, I agree with Cartwright’s analysis. However, by the end of the talk I could not help but to wonder how can the kind of detailed information (the very kind that seems to be required by Cartwright’s view) be acquired and most effectively used by policy makers? Cartwright and I both agree that insofar as Peter’s death threw light on the need for policy reform, we must not let that troubling result rush us into making poor, likely misinformed decisions. Cartwright argued that a unique approach will be required for different locations and different populations. But it does not seem to me that the implications of this aspect of her analysis have been thoroughly explored. It is important to recognize that Cartwright’s approach implies the existence of two distinct groups: The policy makers, and the target populations. Her analysis adds a degree of subtlety that is certainly required. However, I argue that it does not go far enough. Rather than seeing the problem as one of the communication between groups, I urge the reader to try to view the ideal response as a communal project that will result in systematic policy changes. I will call this “the communal approach” and, in general, it suggests a merging of the policy makers and the target population. More colloquially, my idea could be characterized as a grassroots approach.

One of the worries underlying my suggestion is that without a communal approach, any policy alternative will continually have to deal with worry of paternalism. It’s possible that a target population and a group of policy makers could work in tandem to develop and implement some systematic reform. And in this way, they could avoid the spectre of paternalism for a while. However, without a systematically inclusive component to the policy under discussion, to ensure at least the possibility of community involvement, the worry about paternalism cannot be thoroughly addressed. Rather, it will have only been addressed at the level of development and not of implementation.

Paternalism is not only morally problematic, it also presents a pragmatic problem. When a policy is perceived by the general population as not only intrusive, but also excessively controlling they are less likely to want to follow the recommendations of that policy. If a group of people lack the desire to adhere to whatever policy, then it won’t function as effectively as it could have otherwise. Further, no one in the target population has any vested interest in the success of this policy.

Continuing along the lines of pragmaticism, top-down approaches to the development and implementation of some public policy also ignores the worry of trust. The trust of a particular community, especially one that has historically perceived its relationship to government bodies (police, child services, etc.) as an essentially hostile relationship, cannot be taken for granted. There are communities which rightly feel that their interests are directly threatened by a particular public policy.

For example, aboriginal communities in South Dakota have had a long standing complaints against the state’s department of social services, and especially regarding the rate at which their children are taken from their families. These feelings are only intensified when repeated violations of the law go either unreported by the local authorities or unpunished. (Full story here). Even if the best possible reforms were made, it’s not clear that this community would or should believe it. Were they to be involved in the development and implementation of such policies, they might have a reason to. They certainly want to be involved (she the story of Marcella Dion here) and the community driven approach could help to at least foster the development of the relationships required to make any policy changes successful.

We must appreciate, especially in the light of Cartwright’s analysis, that policy makers are at an epistemic disadvantage from the outset. While they have, in the best cases, access to high quality information about a particular population, at the same time, they cannot know if their information is incomplete. Effective policies in one place are not always successful when they are implemented in new areas. The attitudes of local populations towards government assistance, their openness to third party help, and so on will determine the effectiveness of a particular policy when it is implemented in a new location. So it seems reasonable to suppose that giving a community direct input in the creation of new policies would help to not only to draw attention to local problems, but also by building a sense of trust within the community as I have described it.

I do not think that the following conditions constitute an exhaustive list, but I do think that they are a step in the right direction. To begin, we must begin to see both the problems and the people involved as part of one community, where the bounds of our conception of the community are dictated by the nature of the problem. For a local problem, the appropriate community might be a municipality, for example. Secondly, we need to cultivate a policy development strategy that is not unilateral, and that involves a concerted effort on the part of all parties to listen and to discuss in good faith.