Background
The LSE Behavioural Public Policy Knowledge Exchange Group (hereafter the Group), funded by the LSE’s Knowledge Exchange and Impact Fund, was convened to allow those working in the public and private sectors, the international agencies, the media and academia to meet to discuss the role that behavioural science ought to play in informing decisions that affect society.Footnote 1 The Group was not convened merely to advocate for the use of behavioural science in policy. Rather, the Group’s objective was to foster greater appreciation between its members of the constraints, limitations, frustrations and opportunities that they face when attempting to inform policy with behavioural science findings in their various institutional settings.
From its foundation, the Group convened over a biannual one-day workshop hosted at the LSE, with the intention of creating a BPP community of the various interested parties. The first workshop, held on 25 September 2023, was divided into the following four sessions.
Session 1: An outline of the broad initial objectives of the Group, led by Adam Oliver.
Session 2: Public sector perspectives, led by Rupert Gill and Dan Berry.
Session 3: Evidence-policy exchange perspectives, led by Elisabeth Costa and Pete Lunn.
Session 4: International agency and private sector perspectives, led by Chiara Varazzani and Oliver Payne.
I will return to the broad objectives of the Group, which formed the focus of the discussion in Session 1, after briefly summarising the main points presented in the proceeding sessions.
Perspectives on BPP from interested stakeholders
In Session 2, Gill and Berry both alluded to a difference in priorities between academics and public policy makers, in that academic outlets often insist on robustness checks to an extent that serves as an obstruction to the timely production and use of social science for practical policy decision-making. Gill also stated that theoretical knowledge, although not unimportant in public policy, is in much less demand than practical lessons.Footnote 2 Berry felt that academics are overly cautious about making policy recommendations in the absence of an evidence base that is specifically addressed to the problem under consideration, and that behavioural science expertise ought to be leveraged to help diagnose problems and hypothesise possible solutions by interpreting theory and considering evidence from similar settings when such a specific evidence base does not exist.
In Session 3, Costa similarly stated that the use of behavioural science in policy currently suffers from being too mechanistic, which limits its impact. She stated that it should be used more as an upstream lens in strategy formation and problem diagnosis, rather than being largely confined to solutions. Costa also noted that behavioural science applications ought to be more expansive than their tendency to focus on communications, with more choice architecture-style experiments needed. Before running trials, Lunn similarly called for the need to conduct more diagnostic studies that pose such questions as why people behave in the ways that they do, what causes them to do so, and could it be due to behavioural factors or poor economic incentives. Lunn also suggested some pragmatic recommendations, including that behavioural insights in policy work best when the behavioural scientist talks directly to the policy maker, and that formally training civil servants in behavioural science leads to a better use of behavioural-led work in policy, a view endorsed by Gill and Berry.
In Session 4, Varazzani made a parallel point by contending that behavioural scientists in diagnosing a policy issue and tax policy experts in designing structures to improve tax systems need to co-design interventions, policies and systems. On this view, we cannot expect to see a good behavioural intervention if a diagnosis is made and is then simply passed on to the tax team to run with it. Thus, she maintained that there is a need to move away from the mindset that there should be a separate unit of behavioural scientists informing policy and, like Lunn, offered a series of pragmatic recommendations. For example, Varazzani stated that there is a false dichotomy between behavioural public policies and more traditional policies, and if such a distinction is made it is harder to get policy makers to buy-in to behavioural science, which is even harder to secure if policy makers believe that their existing processes are being criticised. She argued that the impression should be given that behavioural public policy is not a niche field; rather, that it is part of business as usual. She also called for a framework that can link individual behaviour change with more systemic change. Finally, Payne emphasised that a core theme for the Group ought to be on the use of multidisciplinary perspectives that go beyond cognitive psychology and economics to delve deeper into why people behave in particular ways in different contexts.
In summary, the following themes and concerns expressed by the speakers at the first of the Group’s workshop were identified:
A. Access to the most technically robust evidence is not necessary to make policy recommendations.
B. There should be a willingness to diagnose the extent to which behavioural science can offer policy insights on the basis of general evidence, rather than to focus upon providing specific evidence-based solutions for particular problems.
C. Policy makers are generally uninterested in theoretical academic arguments, even if theory is central to diagnosing the extent to which behavioural science should be used to address policy challenges.
D. It would be useful for policy practitioners to receive training in behavioural science, which, with academics, would facilitate their co-production of behavioural public policy knowledge.
E. Those advising policy makers invariably find it challenging to convince policy makers of the importance of behavioural insights. Part of this convincing may take the form of illustrating how major policy responses can be influenced by the behavioural affects, although this would need to be done with care so as to avoid the possibility that policy makers may think that their normal practices are being unduly critiqued.Footnote 3
F. Behavioural public policy should be informed by all disciplines that study how and why humans and non-human animals react to their physical environments.
In reaching a consensus on broadly accepted ways forward for behavioural science in policy across the wide spectrum of interests that form the Group, the theoretical arguments underpinning the different approaches to behavioural public policy, perhaps to the chagrin of some policy makers, cannot be avoided. They focus on questions of when, by Group consensus across a range of stakeholders, we can decide that a behavioural public policy intervention is allowable. Note that some of these approaches are deemed acceptable by some but not all members of the Group. The focus of the workshop and of this essay was to identify those frameworks that are acceptable to all included stakeholders. The next section will summarise some of the theoretical arguments, but they will be framed in terms that will, it is hoped, make it clear to policy makers that they relate directly to practical concerns.
Reaching a consensus on the use of behavioural science in policy
As intimated above, in a Group with members from academia and the public and private sectors, among others, there will inevitably be diverse views regarding the appropriate bounds of behavioural public policy. In some respects, their varying objectives may be incommensurable. For example, academics face pressures to publish, civil servants to obey and appease their political masters and the private sector to generate sales and profits. There are also other tensions. For instance, due to political sensitivities, public sector professionals may be reluctant to reveal their views in an open forum that also includes private sector employees, and the latter may be cautious in disclosing their intellectual property. In some respects, those working in the field may be direct opponents – for example, when the private sector strives for ways to circumnavigate government regulations.
There are, therefore, challenges in identifying a consensus in such a Group. The hope was that by having a core Group that meets regularly over an extended period, trust would be strengthened and some of the barriers weakened between the various stakeholders. However, at the outset, an attempt was made to establish some common principles that all members of the Group could accept.
The task was to identify the conceptual behavioural public policy frameworks that all members of the Group felt they could support. Open, non-behavioural-informed education and information provision, such as calorie counts on menus, are of course not among them, not because the Group members do not support this type of intervention but because the design of such interventions are not obviously informed by behavioural science and are thus not part of behavioural public policy. Similarly, the Group members support the use of other traditional interventions, such as financial incentives and regulation, but again, these, when lacking behavioural input, fall outside the realm of behavioural public policy.
Conceptual frameworks that fall within the domain of behavioural public policy must in some way use the findings of behavioural science. These findings are, for the most part, systematic patterns in behaviour that are unaligned with the assumptions that underlie conventional notions of rationality in economics. This manuscript is not the place to summarise these various findings, but for illustrative purposes, one is so-called present bias (i.e. the observation that people place a heavy weight upon the present moment and quickly and heavily discount the future) and another is loss aversion (i.e. the observation that people place a much greater weight on a loss than they do on a gain of the same magnitude).
The conceptual behavioural public policy frameworks can be categorised in a number of ways, but two important distinctions are between those that target internalities rather than externalities, and between those that purport to preserve agency and those that are arguably manipulative or coercive. These distinctions are significant when considering policy acceptance by policy makers and the wider general public. To target internalities is what is meant by paternalism, where manipulative or coercive behaviour change interventions are directed at people for their own good. If an intervention preserves agency it is not paternalistic. Therefore, interventions that are intended to educate or inform people in unbiased, non-behavioural-informed ways are not forms of paternalism because people can ignore the information imparted if they so wish. They remain free to act upon their own agency. The education/information is merely intended to better inform people about the options available to them, without manipulating or coercing people towards particular ends.
Nudge interventions, at least in their original manifestation, were proposed as a soft form of paternalism called libertarian paternalism. The notion was that many people make decisions that harm themselves because they are led astray by the behavioural affects, and thus the policy maker can correct for these errors by reshaping people’s environments such that they are more likely to make better choices for themselves. Crucially – and this is the purported agency-preserving aspect of nudges – it was proposed that when faced with this sort of intervention people do not have to change their behaviours if they do not wish to do so. However, by relying on the behavioural affects that drive automatic decision-making for their impact, the underlying motivational force underlying this type of intervention is inexplicit for most of the people that they target. If these interventions move people’s behaviours in particular directions without people realising that they are designed to do so, one may argue that they are inherently manipulative, even if the new behaviours are ones that the target audience might prefer (which is an eventuality that is by no means guaranteed).
In the workshop discussions, some members of the Group expressed support for the soft paternalistic approach. They contended that it is legitimate for policy makers to attempt to improve people’s wellbeing in this manner. Other members of the Group believe that such tactics are inconsistent with the rules of deliberative democracy, and that, moreover, policy makers cannot hope to understand what individuals want from their own lives, given that inter and intrapersonal desires are multifarious. Therefore, the members of the Group could not reach agreement on the legitimacy of soft paternalistic interventions. As such, there was no Group consensus in supporting the use of soft – or hard – paternalistic behavioural public policy interventions.Footnote 4
However, non-manipulative, non-coercive behavioural public policy frameworks exist. One such framework has a number of arms, including that of designing choices in ways that humans are most easily able to process – for example, for risky decisions, options may be presented in terms of natural frequencies rather than probabilities. Another arm of this approach involves educating people about the various behavioural affects so that they are better equipped to guard against being manipulated by other interests, and so that they may choose to modify their environments as a means of controlling their own behaviours if they so wish – for example, by placing supplies of chocolate at the back of kitchen cabinets to curb temptation. A conceptually related approach advocates that all soft paternalistic interventions be accompanied with a clear explanation of how the behavioural affects inform their design and for what purpose, thus dampening the charge of manipulation. The frameworks summarised in this paragraph perhaps rely for their effectiveness on a lot of mental effort by their target populations, and in many circumstances the expectation that this effort is forthcoming, at least in a sustained sense, may prove unrealistic. However, both approaches can be thought of as forms of behavioural-informed, or at least behavioural accompanied, education/information. There was consensus within the Group that these forms of intervention are acceptable.
Other behavioural public policy frameworks address externalities – i.e. the harms or forgone benefits that people impose on others. One of these forms has the same characteristics as soft paternalism, with the exception that the target outcome is the mitigation of external rather than internal harms.Footnote 5 For example, a policy maker may use the behavioural affects to design an intervention that, in essence, manipulates the target population so that they are more likely to engage in climate friendly behaviours on the justification that their current climate unfriendly behaviours are potentially harming people other than themselves. The Group members deemed attempts to mitigate harms in this way as acceptable.
There are two ways in which regulations against externalities may be justified on behavioural grounds. One of these mirrors coercive paternalism, but with a focus on externalities rather than internalities. That is, one party may be influenced by the behavioural affects to act in a way that imposes harms on others, to the extent that those actions could be subject to penalties. For example, the managers of a manufacturing plant might sanction the pollution of a river due to paying insufficient attention to the future negative consequences of their actions, if regulations are not in place to ban those types of activity. However, the legitimacy of regulations against such harms to third parties does not typically require input from the behavioural sciences. Regulation against polluting rivers for personal gain is likely to attract broad support irrespective of whether the actions of the polluter are behavioural or, from their own perspective, rational.
The second form of behavioural-informed regulations are potentially introduced when one party to an exchange uses the behavioural affects to manipulate another party to do something that they would not otherwise have done, and it is deemed that the manipulation has unacceptably harmed the manipulated party. For example, gambling companies might offer a range of daily maximum spend limits that their consumers can commit to, but many consumers may anchor on the highest limit on offer if its placement is salient, which could induce them to spend more than they would otherwise have chosen to spend. If it is concluded that the companies are using behavioural-informed processes that impose unacceptable harms on people, then there is a legitimate argument to at least consider regulating against their practices. The members of the Group supported using behavioural-informed regulations to mitigate harms to second and third parties.
The Group considered one further approach to behavioural public policy that applies to broader institutional rules than the frameworks discussed above and concerns the almost instinctive motivation for people to cooperate with one another. Given the right laws and institutions, humans are natural social cooperators, but these motivations can easily be eroded and replaced with more egoistic tendencies if the institutions are not supportive of them. A challenge for behavioural public policy, which the Group accepted, is to try to ensure that the broad institutional rules and structure of society facilitate social cooperation.
To summarise, the Group members could not reach agreement on whether behavioural public policy should embrace paternalistic approaches. The one exception was where the members potentially supported paternalistic interventions that are explicit (i.e. in the form of a regulation, rule or law that can be openly debated) and impose minimal constraints on each individual, and yet potentially reap large societal benefits. Possible examples are laws that enforce the wearing of seatbelts and motorcycle crash helmets. Not all members of the Group are entirely comfortable with even these instruments of paternalism but are willing to accept them.
The Group did reach consensus in supporting frameworks that aim to retain individual agency over decisions that affect principally those exposed directly to behavioural interventions, and frameworks that are intended to mitigate harms imposed on second or third parties by a first party’s actions. These frameworks include:
A. Using knowledge of the behavioural insights to make it easier for people to understand the decisions they face.
B. Educating people about the behavioural insights so that they can modify their own behaviours if they so wish.
C. Providing an explicit explanation for the motivation for any intervention that may otherwise be viewed as manipulative and paternalistic next to the intervention itself, so that people remain free to act upon their own agency.
D. Using behavioural-informed manipulations of individual behaviours in those cases where they mitigate harms or enhance benefits to/for people other than those who are manipulated.
E. Regulating against actions caused by the behavioural affects that potentially impact negatively on third parties.
F. Regulating against behavioural-informed manipulations by first parties to an exchange that potentially impact negatively on second parties.
G. Regulating and legislating to protect and enhance pro-social instincts.
The above intervention types are thus those that the Group members agreed merit most attention in that these are likely to be met with broad acceptance. Some members of the Group were more accepting than others of paternalistic interventions, but this essay is intended as a consensus statement that places emphasis on only those frameworks that all Group members would accept. The accepted frameworks still encompass a wide range of behavioural public policy interventions across every conceivable domain of public and private policy. This does not mean that all interventions that meet the criteria of any of the above frameworks would be deemed acceptable to policy makers or the public. Each intervention would need to be considered on a case-by-case basis to analyse its potential costs, benefits, the balance of harm mitigation versus restrictions on freedom, unintended consequences, and other political and ethical objections. It means only that the intervention meets the conceptual requirements of a behavioural public policy framework that is consensually accepted by the Group members.
To reiterate and to conclude, the members of the Group reached common consensus in supporting the use of behavioural public policy to strengthen – or at least retain – individual agency in the decisions that people take that affect their own lives, to target externality concerns and to protect and nurture the social instincts.
Open access
