Last night I had the pleasure of attending a reception in honor of Heiner Bielefeldt, Special Rapporteur on Freedom of Religion or Belief for the United Nations. Bielefeldt recently released a report about relgious freedom in the workplace and the reception was a chance for him to talk about that report in a less formal setting as well as to hear commentary from a variety of interested parties.

It was fun. I used to attend these sorts of things all the time. Back when I was doing strategic and policy analysis for the U.S. Department of the Army, I regularly attended working groups at the RAND Corporation or gave presentation at National Defense University. I was used to the policy-wonk/quasi-academic world of “let’s all get together and talk about how to solve a hard problem.” However, I left that life several years ago. A tweet from Nasim Taleb today conveyed a little of what it feels like to jump back in for a night:

It wasn’t quite that alarming, but I definitely did feel out of my element, which made me feel even stranger since I used to navigate or even take the stage at these types of functions on a regular basis. Since I left my work with the Army, I’ve focused on technical implementation - building systems that make or augment decisions, and creating processes to ensure the smooth functioning of those systems. I used to try to solve a few big problems in general. Now I work to solve lots of little problems in particular. Last night reminded me just how different those two approaches can be.

People sitting near me probably thought I was texting throughout the whole discussion, but I was actually furiously taking notes. The event made me see policy in a way that I’d never thought about it before, and I needed to write down a bunch of reminders of what that looked like because I knew I wouldn’t have a clear view by the next morning. This post is mostly an attempt to turn my list of notes into something more coherent.

If you go to the press release announcing Bielefeldt’s report, you read: “Mr. Bielefeldt’s report offers a number of practical recommendations, and puts forward the concept of ‘reasonable accommodation’ as a tool to put into context the principle of non-discrimination in such a way that appropriate individual solutions for religious minorities can be found at the workplace.”

That except from the press release illustrates the problem I see with a lot of policymaking. When we talk about “religious freedom in the workplace,” we’re actually talking about a whole lot of different things. We’re talking about “wearing a piece of clothing” (such as a head scarf) and “being afforded time and a private space to participate in a ritual” (such as prayer) and “not having to work on a particular day” (such as a religious holiday). Those are all different behaviors that may give rise to disputes that require very different courses of action in order to reach a resolution. When we treat them all as one “issue” we have to come up with general (philosophical, theoretical, legalistic) ways to talk about it, which makes it hard to come up with anything more than very general solutions. It seems that it’s necessary to come up with a whole concept of “reasonable accommodation,” a term that itself requires familiarity and perhaps even training to use properly, exactly because we have this term of “religious freedom” that doesn’t point to any kind of concrete behavior.

I ran into this working for the Army. People would talk about the need to fight corruption in Afghanistan. But no one ever pointed at someone and said “look, that person is doing corruption.” Likewise, we can’t point at someone and say “look, that person is religious-freedom-ing.” What we do see is someone taking money in exchange for political favors or refusing to work on a religious holiday. It seems to me that “religious freedom” and even “religious freedom in the workplace” are too separated from actual observable behaviors to be a good focus for policy. The concepts are too abstract.

This is a practical concern: generalized problems give rise to general solutions. General solutions are hard to implement and often don’t work well. We just end up arguing about the meanings of particular words instead of building stuff that makes things better. Discussions about general solutions are easy to derail because they are tangentially related to so many other general but emotionally-charged topics that they can be attacked, diverted, or obscured with just a few words.

All of this can be clearly seen as soon as a policy discussion switches from the topic of what to do, to how to do it. “We need more training and greater understanding” is practically a trope. It assumes people do bad stuff because they don’t know any better, which is both patronizing and of questionable accuracy. But that assumption allows us to focus on something we know how to do - make rules about how people are supposed to act and then talk to them a lot about how they are supposed to follow the rules - rather than the less familiar but possibly more effective task of creating a good reason for people to do what we want them to do and then making it easier for them to actually do it.

Policy largely deals with answers that aren’t supposed to change: when you set a rule (policy) in place, the idea is that that rule should stand for a good amount of time with as little modification as possible. Policies run into trouble when they encounter scenarios that weren’t anticipated when the original policy was formulated. For example, Bielefeldt’s remarks focused entirely on situations where an employee wanted to practice some sort of religious observance and the employer resisted that. A couple audience members asked questions about the reverse: what about an employer that feels a religious injunction against paying for something (such as employee’s access to contraceptives), even though the employees don’t share that religious perspective? What if an employee, from the employer’s perspective, is currently experiencing or exists in a state of religious impurity based on caste or some other designation? Bielefeldt’s response, it seemed to me, was to wave off those concerns as largely inconsequential, as individuals should not be able to use their personal religious beliefs to dictate or limit the behavior of their employees. From a policy perspective, that’s a reasonable enough position, I guess. However, so is the United States Supreme Court’s majority opinion, which disagrees with Bielefeldt.

The problem I have with both positions is that they are unnecessarily adversarial. They set up a situation where two people’s intended behaviors - the employer’s and the employee’s - are in conflict, and where there will have to be someone who wins more (perhaps all) and someone who wins less (perhaps nothing). It doesn’t have to be that way.

Let’s pull this down from the lofty heights of “religious freedom in the workplace” and focus on the actual behaviors. An employer pays for insurance for employees, but does not want to pay for access to contraceptives. An employee works in exchange for, among other things, access to insurance, and wants that to include access to contraceptives. How do we resolve that tension? First, have the insurance companies figure out a tiered pricing system - the price of a plan with contraceptive coverage and the price of a plan without contraceptive coverage. Second, build a platform - say, a website - where people can register a need for the difference between the contraceptive-included and contraceptive-excluded coverage to be paid for. Allow people to donate through the website. Make a paypal-like transaction: if you feel that people should be allowed to have access to contraceptives through their insurance, donate some money. That money will be directly allocated to the plans of people who have said they need that coverage. Employees get contraceptive coverage. Employers don’t pay for it if they’re relgiously opposed to it. Everyone wins.

At this point, I wouldn’t blame you for thinking that I’ve ignored all the many, many difficulties that would be inherent in that plan. How do we get insurance companies to create a tiered system? How do we avoid a future where they have to create a million fine gradations in their coverage for every little exception people want to make to every rule? How do we keep all employers everywhere from just dumping their contraceptive coverage so they don’t have to pay for it? How do we ensure the security and appropriate level of anonymity in the proposed system for channeling donor dollars to applicants? What do we do if we can’t get enough donors for the system to cover everyone’s needs?

Two responses:

  1. A lot of those problems actually have policy solutions. For example: only require coverage tiers when not doing so would violate a constitutional right. Publicly post the list of employers that opt out of contraceptive coverage - those who do it for religious reasons will see it as a badge of honor and those who do it because they’re cheap will be shamed. All of those problems are smaller and more manageable than the problem of “how do we avoid religious coercion in the workplace?”

  2. Think of how much time, effort, and money it would take to make that kind of system. Now think of how much time, money, and effort it takes to navigate the U.S. legal system. I’m not saying my suggestion is easy. I’m saying the way we do it now is as least as hard, and the way we do it now results in one or both of the parties feeling pretty aggrieved, which seems like a pretty fragile solution.

I’m talking about implementing a system solution: instead of trying to come up with a set of rules regarding high-level aggregates of lots of different behaviors that cause conflict, create applications that address each of those behaviors individually, preferably by allowing both parties in question to get what they want. System solutions face technical barriers, not moral or philosophical ones.

Several years ago, Wired came out with an article with the dumb title “The End of Theory: The Data Deluge Makes the Scientific Method Obsolete” The article didn’t really argue that data kills theory, of course, but in its science-journalism-y way did explore, among other things, the point that having a constant stream of actionable information does limit the extent to which we really need grand, over-arching ideas to guide our behavior.

That’s a system solution: engineer the capability for people who have a need to create a constant stream of information about that need, and direct that stream to people who have the desire and ability to fulfill that need. Data enables rapid adaptation. You don’t need a general policy rules if you have the data to accommodate exceptions. The whole concept of an ‘exception to the rule” loses a lot of its meaning in the context of an adaptive system. As an added benefit, if we iterate solutions we don’t have to worry as much about “getting it right” as we do if we’re trying to formulate a general rule that can last with minimal modification.

Obviously a system approach can’t and shouldn’t be applied to all policy issues immediately. It’s highly dependent upon good data collection. The toy example I gave above regarding contraceptive coverage is a relatively straightforward problem because the problem is easy to measure - the need is literally measured in dollars and cents. Other problems aren’t so standardized. And there’s certainly still a strong need for all those people who do policy: systems still have to function within the policies that currently exist, or else those policies need to be amended to accommodate the new systems.

In general, I prefer to frame problems as system problems rather than policy problems because doing so forces us to talk about behaviors and outcomes, not theories and ideas. I know I’m biased, but it just seems a whole lot more practical. It seemed strange last night hearing people talking about “effecting change on the ground” on the basis of a static report. That report is obsolete the moment any real changes are made, because the challenges that prompted it will have been altered, even though they will almost certainly not have been eliminated. I don’t think data heralds the end of policy any more than it heralds the end of theory, but I do think a lot of what we currently consider policy problems don’t really require policy solutions.

If you want to comment on this post, write it up somewhere public and let us know - we'll join in. If you have a question, please email the author at