There follows a guest post by former Google software engineer Mike Hearn.
How can we avoid a repeat of the last two years?
To ensure policy failure on such a scale never happens again, those of us who oppose them need concrete legislative proposals that could be implemented by a parliament or congress, and which address the root causes of the failed policies themselves. Very often in history we see that ideas for political reform have to be kicked around the public sphere for a while before being picked up by politicians. In that spirit I lay out some proposed changes to the law, designed to encode lessons learned from the Covid pandemic. Not all of these proposals apply to every country and they take for granted the acceptance of a viewpoint that is still contested – namely, that Covid non-pharmaceutical interventions (NPIs) were a mistake. But the ideas here will hopefully prove useful as a launching point for further discussion – and perhaps, eventually, political campaigns.
My goal here is to make proposals that are only partially within the Overton Window of currently acceptable political thought. The justification: ideas fully within the Window will be generated by politicians during any normal public inquiry anyway. Ideas fully outside it won’t be considered at all. All proposals should be somewhat uncomfortable to read for someone fully committed to mainstream politics, but not entirely so. Please note that anything related to pharmaceutical or financial interventions are out of scope for this article. Further work (perhaps by other people) may address legislative proposals around these.
Repeal all Emergency Acts
Many countries have one or more Acts which grant ministers near arbitrary powers in the event that they declare a state of emergency. These Acts have a history of abuse without providing concomitant benefits.
The U.K. Public Health Act 1984 (PHA) has little or no history of use before Covid. NPIs were implemented primarily using the PHA and the Coronavirus Acts. However, NPIs were not effective, and their implementation led directly to many negative outcomes including severe acts of injustice against citizens and levels of government spending that have triggered an inflationary cost-of-living crisis. In the case of Canada, the Emergencies Act 1988 has only ever been invoked once, and that was to end what was otherwise a lawful protest. That Act replaced the previous War Powers Emergency Act after it was heavily criticised by the McDonald Commission, due to another case of abuse: Justin Trudeau’s father invoked it during a peacetime kidnapping crisis.
Modifications to Emergency Acts are thus well within the political mainstream and in fact a proposed reform of the U.K. Public Health Act has already been tabled in Parliament. Politicians seeking the centre ground often prefer to think in terms of reform rather than repeal. However, I think the case for a simple repeal of all such laws is a strong one, along with constitutionally forbidding the creation of new Emergency Acts in countries that have a constitution.
The argument for full repeal is as follows.
Prior reforms haven’t worked. The Canadian experience is informative. After their Wartime Powers Act was abused to intervene in a non-military crisis they replaced it with a reformed Act. But after decades of non-use the successor was invoked for the first time only to eliminate political dissent that could have been easily handled in other ways. Thus the Canadian Emergency Acts have never provided the country with any clear benefits, only international embarrassment.
Emergency Acts are unfixable by nature. Because of the open-ended nature of Emergency Acts, the only means available to prevent abuse are either to narrow the scope of what can be considered an ’emergency’, or to narrow the scope of what is allowed during an emergency. However, by its very nature an emergency is defined as a situation in which normal oversight mechanisms are too slow. Thus, by the time a Parliament, Congress or court is able to debate whether the powers have been invoked correctly it’s already too late. Once again, the Canadian experience is informative. Trudeau’s powers expired after 30 days, but this was more than enough time to complete the process of crushing his political opponents.
The underlying intuition is incorrect. The justification for Emergency Acts is that while the safeguards and procedures developed over hundreds of years by democratic societies lead to slower but better lawmaking, there are times when making law fast is better than making law well. For this justification to be valid requires several things to happen at once:
- The situation of an entire country or society must change overnight in ways that nobody could have anticipated or planned for.
- A response is required that only a government can implement.
- That response is nonetheless required only temporarily (usually declared to be a matter of weeks).
- The Prime Minister or President is able to both immediately determine what that response should be, and correctly implement this plan without making any major mistakes, even though in non-emergency times this isn’t really possible (which is why Parliaments developed alongside monarchies in the first place).
- For undefined reasons the populace refuses to agree that the government response is correct, and must be forced to comply by violent coercion. If this condition wasn’t met then no emergency powers would be needed because the government could simply issue advice.
The problem is, the set of situations in which all these conditions apply simultaneously is the empty set. There are no cases in which a temporary, correct and yet coercive response that only governments can mount is required for a completely unanticipated situation. It’s for this reason that Emergency Acts often go decades without being used after they are passed, and when they finally are used they are often reformed afterwards.
Covid fails at least requirement 1, given that global pandemics were widely anticipated by the public health community to the point of misfiring (e.g. for Swine Flu). There should have been no need for new law. It’s plausible that the open-ended nature of the Public Health Act actually discouraged Parliament from writing higher quality law, under the impression that what to do in a pandemic was a solved problem already. In the beginning it would also have failed requirement 5, given that lockdowns were based on widely held intuitions about how respiratory viruses spread, and thus voluntary compliance in the population was very high. Finally, given the widely discussed problem of expert failure during the pandemic it also failed requirement 4.
Emergency powers are often claimed to be necessary for fighting wars. Again, we can see that they should not be required for these situations. Nothing about war invalidates the requirements above and nothing in war is likely to ever meet them. War is not only something that can be anticipated; anticipating it via wargaming is a key activity of any competent military. War rarely starts overnight without warning but even when it does governments have either pre-prepared via defence spending and thus already have the necessary laws debated and on the books via the usual processes, or they’re unprepared and thus about to be defeated anyway, in which case questions of law are irrelevant.
Conclusion: there are no situations in which the underlying philosophical justification for emergency powers are valid, and they are prone to abuse. Therefore they should be repealed.
2. A legal requirement for pre-publication of information presented to the government
Establishing the principle that policy should never be defined in response to a (real or claimed) emergency enables other critical reforms.
Freedom of Information Acts are one of the great achievements of liberal democracy. By converting the default state of government information to public rather than private, FOI Acts open up government decision-making to scrutiny by journalists, the wider public and the members of Parliament who can be bothered to read them. The culture of open data that follows along with these acts is why it’s possible for the public to produce and consume sophisticated data analysis, often revealing facts that governments find embarrassing.
Despite the greatness of FOI Acts, they can be improved. One of their biggest weaknesses is their retroactive nature. Someone must surmise that information exists, explicitly request it and then wait for a response. By the time this response arrives it may already be too late. A good example of this is how the UK FOIA was used to force the release of the source code to the Ferguson/ICL Covid model, revealing that it was so buggy that it wasn’t fit for purpose. But by claiming they were in an extended, long drawn out process of releasing it, ICL was able to delay the opening of the code until months after it had already influenced government policy. By the time the model’s inadequacy was revealed it was too late.
A very similar second situation occurred when sceptical doctors were invited into No.10 in 2020 to argue against lockdown as a so-called red team. SAGE presented new modelling that the red team hadn’t previously seen and thus was unable to immediately respond to. SAGE therefore won by default and presented the new models at a press conference just hours later, leading ministers to claim they were “bounced” into another lockdown. The modelling was shown to be nonsensical just two days later by one of the invited scientists (Dr. Carl Heneghan), but by then it was too late and the decision had already been made.
If the principle that governments must sometimes change the law within hours is successfully discredited, that enables FOI Acts to be strengthened by requiring publication of any data or information before it is presented to ministers or civil servants by third parties. A suggested change could be as follows:
- The government is mandated to run a website on which all documents, presentations, data sets and supporting research artefacts to be presented as part of any government meeting or event to government employees can be uploaded, downloaded and commented upon.
- No externally generated document, presentation or dataset may be viewed by any minister or civil servant until it has been available via that website for a minimum of three days beforehand.
- This requirement does not apply to:
- Internally generated emails, e.g. between ministers or civil servants.
- Documents that are generated regularly according to a pre-agreed schedule, e.g. status reports, dashboards, etc.
- Contracts, commercially sensitive information, etc.
- These sorts of exceptions would undoubtably be abused immediately, but starting with broad exceptions provides a foundation that can be further refined.
- Presentation of information to the government before the pre-disclosure period has expired is subject to fines. ‘Presentation’ is defined as including provisioning of the document in any form, including links to the documents on the document collection website (i.e., ministers and civil servants should always be the last to know what’s about to be presented to them).
There would be no requirement for anyone in government to actually pay attention to any resulting comments, only to provide infrastructure for collecting them. The primary burden would be regulatory compliance: anyone interacting with the government would have to be prepared enough to write their PowerPoint slides at least a few days in advance, and in large institutions that interact with the government regularly that would probably require new compliance controls to be put in place.
On the other hand, for government agencies themselves it may actually save money in the long run because pre-publication would largely eliminate the need for dedicated staff to handle FOI requests. In a few cases government departments have already realised this and published everything they have on frequently FOIAd topics, e.g. the U.K. Ministry of Defence stated it opened up its UFO report archives specifically because it was easier than responding to FOIA requests individually.
During review, a common objection to this part was that it would cause advice to go underground and unwritten, or the compliance burden would cause people to simply not present evidence to the government at all, or to significantly change that advice due to fear of backlash. I don’t feel like this is a major problem because Freedom of Information Acts already establish the principle that any advice or written documents given to anyone in government can be made public anyway, and for important issues probably will, so that bridge was crossed long ago. Also, if freedom of information actually does cause people to moderate their advice to governments, then we haven’t seen much evidence of this during the Covid pandemic. SAGE scientists were more than happy to repeatedly go to the press to make their views even more public than what was already being achieved by the release of their meeting minutes.
3. Establish a Parliamentary Scientific Methods committee
A major weakness exposed by the Covid response is that governments want to “follow the science” but lack any formal definition of what science actually is. This has been repeatedly exploited by academics who present themselves to ministers and the public as scientists, yet whose work doesn’t meet any standard definition of the scientific method.
The root cause of this problem can be traced to the grant approvals process. Governments treat the necessity of spending of money on research as a matter of bipartisan agreement, so there’s little incentive to ensure the money is spent well. Grant money is dispersed with little regard for whether the resulting papers correctly utilise the scientific method or even if they have anything to do with the original grant proposal at all.
This approach is killing the goose that lays the golden eggs. By flooding the research world with undiscerning ‘dumb money’, governments have been incentivising the production of cheap but bad research papers that not only pollute the scientific literature, but which starve good work of oxygen. The fixation on unvalidated modelling (cheap to produce and in endless supply) would appear to be an example of this.
Fixing it requires governments to start caring more about how exactly public research budgets are used. In turn this requires creating a rigorous definition of what is and is not scientific. Existing scientific institutions are sadly of little use here:
- Scientific journals define science as anything which is both interesting enough to publish and which has passed peer review. This definition can sometimes catch individual papers with low standards, but can’t arrest a general decline across an entire field.
- Universities appear to define science as anything that grant-making bodies are willing to fund, or journals are willing to publish. From an outsider’s perspective they really couldn’t care less about the credibility of their academics.
- The commercial sector is also of no use in defining science, because industrial labs are typically focused on measurable results rather than the philosophical question of what is or is not scientific. The question arises in the public context only because:
- Governments force people to pay for basic scientific research on the grounds that it’s a public good that would otherwise not be funded by anyone at all – thus deciding what is or isn’t scientific is about responsible use of taxpayer funding.
- Governments force people to change their lives in major ways on the grounds that ‘science’ says it must be done – thus deciding what is or isn’t scientific is also about basic civil liberties and quality of governance.
Wikipedia’s article on science defines it as follows:
Here are just a few examples of common cases where publicly funded research doesn’t appear to meet this definition of scientific:
- The production of models which are never tested for predictive validity, or in which the predictions cannot actually be tested at all due to lack of falsifiability, determinism or clarity (not testable).
- Claims in psychology that don’t replicate (non-replicating claims aren’t knowledge).
- Papers with logical fallacies in them (not systematic).
- Fraudulent practices like making up data or Photoshopping images (not building true knowledge).
- The practice of presenting probability-free scenarios as science (not testable).
- And I would be remiss if I didn’t take a pop at my own field of ‘Computer Science’ (not about the study of nature).
Many of these points are obvious. In theory, nobody should claim that a fraudulent paper is scientific, except that our current system actually does so and it happens all the time. Therefore the first step to raising standards is to write down why they are presently unsatisfactory.
Because it’s ultimately about tax and spending, this is a task for Parliament. Therefore, a standing committee could be set up tasked specifically with creating and maintaining a legal definition of the scientific method, which granting agencies are obliged to implement and audit. The committee could meet on an occasional but regular schedule and take evidence from the public on cases where the definition doesn’t seem to be correct; either because it’s allowing unscientific work to be funded, or because it’s incorrectly excluding a new area of research that should be considered science. For example such a committee could define national standards around the threshold for something to be considered statistically significant. Although such a committee could, should and would receive wide ranging input from many different people and organizations about what is or is not scientific, only people outside the system have any hope of imposing a working definition: people within the public sector research sector have visibly failed at imposing standards on themselves.
For the correct lessons to be learned from the pandemic, actionable ideas are required.
These ideas should fall on the border or just outside of the window of currently acceptable thought because, by definition, people within the system will already come up with the ideas that are considered acceptable to propose. Additionally, the sort of public health workers who came up with lockdowns, mandatory masking and so on have demonstrated a willingness to go well outside the window of previously acceptable ideas. Indeed, they entirely redefined the window almost overnight, so the importance of ‘reasonableness’ or ‘acceptability’ as political concepts is open to question anyway. It certainly didn’t stop anyone in SAGE.
I argue for three actionable ideas: repeal of emergency laws that suspend the usual decision making process, changes to the way information is presented to ministers, and an attack on public sector pseudo-science via reform of the granting process to include a rigorous definition of the scientific method. Combined these would have blocked large amounts of the output of computational epidemiology (because it’s unvalidated and thus unscientific), slowed down the initial response (which in the end didn’t matter anyway), and given time for sceptical views to be fully presented to decision makers.
The author would like to thank Toby Young, MTF and Harry Richer for their review.