Nov 07, 2018
Lately there has been considerable concern among effective altruists about burnout. People are worried about themselves or others being less productive* or just plain miserable. The goal of this report is to bring people up to speed on the scientific research about burnout and, when possible, make recommendations about alleviation and prevention. Unfortunately, the scientific literature has few specific recommendations to make, so I would like to use this as an opportunity to foster discussion about what has worked and not worked for people personally. Look for those comments below. The goal is for this to be useful to individual workers in treating their own burnout, and to organizational decision makers in preventing burnout organization-wide.
[*Studies are actually mixed on if burnout reduces productivity, with some even showing burnout associated with higher productivity. My interpretation here is that high standards for yourself lead to both high performance and burnout.]
The official definition of burnout is “physical or mental collapse caused by overwork or stress”. That kind of implies that a person can’t work when burnt out, but that’s not my experience- ceasing work when you’re burnt out is a privilege. But working when you’re burnt out is miserable, and makes burnout worse, so even if circumstances improve you’re in a hole.
Burnout was originally conceived of in the caring professions (e.g. nursing and social work), which are emotionally demanding in several different ways. This has by and large not been born out scientifically; other professions burn out just as hard, with perhaps slightly different patterns on the Maslach Burnout Inventory, the most popular measure of burnout. Based mostly on personal observation I strongly suspect there are multiple types of burnout, which can co-occur, and which current instruments are not sensitive enough to differentiate. Of particular interest to this crowd is the difference between burnout caused by hating your job or not having the resources it demands, vs. loving your job too much and being sucked into giving more than you should. I suspect that the latter is more heavily represented among effective altruists than in the literature.
The Maslach Burnout Inventory (used in >90% of studies) divides burnout into three parts: exhaustion, cynicism, and (perceived) personal efficacy. The MBI has been shown to be internally consistent and cross-culturally valid. On the other hand, it has mixed results in distinguishing burnout from traditional depression or anxiety, and I could find no studies demonstrating any predictive value of the inventory — the closest was two studies showing MBI predicted an increase in thoughts of suicide and dropping out of school among med students.
In contrast, the Copenhagen Burnout Inventory has one whole study showing a that a high score predicts future sickness absence, sleep problems, and use of painkillers. The CBI measures only exhaustion, and separately tracks personal burnout, work burnout, and client burnout. I would have liked to give preference to studies using the CBI because it has more empirical validation, but there simply weren’t enough to rely on, so most of the studies referred to in this post use the MBI.
By far the most popular model of burnout in the literature is “Job Demands - Resources”, or “JD-R”, which posits that high job demands lead to exhaustion, and low resources lead to cynicism and feelings of low personal efficacy. “Demands” and “Resources” are defined fairly broadly here. Demands includes things like “coping with conflicting goals” and resources includes things like chances for advancement.
There is a similar theory called Conservation of Resources. It has less empirical support and does not make noticeably different predictions, so I won’t pay it further attention.
An unfortunate fact is that the more privileged you are, the easier it is to create workplace that accommodates you. These suggestions will be easier to implement for people in higher demand professions, who earn more money, or have better bosses.
A second unfortunate fact is that the literature was not very helpful in recommending ways to prevent or relieve burnout. The following are either based on interpretations of literature (e.g. because social support is correlated with low burnout, I recommend social support) or my own beliefs based on what I’ve observed personally.
I’ll create threads for all of these seeking anecdotes about how you’ve achieved them for yourself in the comments.
Unsurprisingly, sleep duration is negatively correlated with burnout. I think the causality here is up for debate — stress causes poor sleep in parallel to poor sleep causing depression — but to the extent you can improve your sleep, it’s likely to be helpful.
This is a more general application of “don’t commute”.
Of course, this advice is only actionable if you have money. However there is a noticeable contingent within EA of people pushing themselves to live on much, much less than they earn, and it’s worth examining whether this is penny-wise-pound-foolish for you personally. Buying rest might reduce your donations in the short term, but if it increases your long term capacity, it is the right thing to do.
80,000 hours goes into the multitude of reasons you should do this here, but I want to explicitly call out savings’ role in fighting burnout. Knowing you can leave can enable you to demand better treatment. Knowing you don’t need a particular job lets you hold out for a better fit.
Vacation is commonly considered the treatment for burnout, however one of the most consistent findings in my research is that the reduction in burnout following a vacation dissipates very quickly — in less than three weeks.
This is one of those times I feel a struggle between what I “know” anecdotally, what research demonstrates, and my personal experience. I’ve seen people in very demanding jobs go on vacation, clear their head, and come back thinking more strategically. But this has never worked for me personally: when a job is unpleasant, it’s unpleasant, and leaving for a week doesn’t make it more pleasant. I conjecture that there are really two forms of burnout — “this job is terrible” and “this great job is very demanding”, and there simply aren’t enough of the latter to show up in studies. This is unfortunate, given that I expect it to be overrepresented among people reading this. But at a minimum, you should not count on vacation to fix a bad work environment.
Ultimately burnout is prevented by individuals, not companies. The warmest, fuzziest company in the world can’t prevent people from working themselves into burnout if they’re truly determined. At the same time, organizations are setting the incentive structures that determine whether preventing burnout is an uphill fight or not, and I think this makes them the correct place for most interventions.
Unfortunately nothing I read went into what constituted good feedback or how to give it. However in general reducing ambiguity is good and reduces burnout.
It has become popular to set “stretch” goals. Google’s quarterly evaluation process says that if you regularly achieve more than 60-70% of your goals, you’re aiming too low. This has the advantage of never leaving someone under capacity, but also means they never get to feel like they’ve “won”, which leads to burnout.
As a bonus, keeping goals modest leaves people with enough reserves and capacity to sprint when you really need it.
Feeling socially supported is one of the few demonstrated protectors against burnout. The literature is mixed at describing specific interventions — some studied deliberate interventions like peer support groups, others merely counted the number of positive and negative interactions. I report the peer review supported interventions below, and leave the anecdotes to the comments.
A warning: workers often recognize attempts to facilitate group cohesion and feel obliged to fake feelings of connection when they can’t feel it authentically. My belief is that this demand for emotional labor can make burnout worse. You can mitigate this by providing a bonding budget for the employees themselves choose how to spend — although even then, not everyone will love the same things, and people may feel pressured to participate or left out if they don’t. You can also make explicit in the interview what kind of social expectations you have, and let people who aren’t compatible opt out.
In a result that is very gratifying to me personally, remote work and flex time generally reduce burnout. However it can’t be done blindly. My personal experience has been that if a company’s workflow is set up for on-site work (e.g. no one responds to IM and you have to physically track people down to get answers, information is generally not written down), remote work reduces personal efficacy to the point that it’s not worth it. But when it is supported (meaning, co-workers are responsive, a strong culture of writing things down), telecommuting removes commuting, reduces work-life conflict, and gives people more control over their environment.
As a bonus, remote work makes it impossible to know how many hours people are working, removing the incentive for productivity theater.
Two caveats to this:
I found no peer reviewed study on this, but firmly believe nap rooms to be helpful when properly implemented. A note to Google: in a busy hallway next to the noisy massage chair does not count as proper implementation.
Especially in international companies, people can be pressured to attend meetings both very early and very late. I don’t have a good solution to this as long as time zones are real.
At this point the evidence is overwhelming that most people hate, are less productive and more sick in open offices. Open offices don’t even increase in person interactions between co-workers. Plus you can nap in private offices.
Autonomy can conflict somewhat with having a clearly defined role and goal, and the correct balance will vary across people.
If you measure hours visibly at work, people will Goodhart and put in more hours at the office. This may not make them more productive, and depending on who you ask may not be able to make them more productive- humans only have so many hours of thought-work in them per day. But it will eat up their time and prevent them from recovering.
Like a lot of young organizations, EA orgs tend to have managers who were hired for things other than management, have not received explicit management training, and have priorities other than management. This leads to a lot of predictable problems- they don’t delegate well, don’t give the right feedback at the right times, etc. “Learning to manage” would lead to more work accomplished and a better work environment for everyone. Unfortunately everyone else is making this mistake too, and good resources for learning to manage are thin on the ground. But a good start would be to simply recognize management as what you are doing and a skill that needs to be learned.
There were a number of topics I hoped to cover but couldn’t, due to a lack of data.
Thanks to CEA for commissioning and funding this research.
If you are exceptionally curious about my research, you can view my notes here.