by Brian Tomasik
First written: 16 Dec. 2016; last update: 23 Jan. 2017


This post describes a few of my personal thoughts and experiences regarding the slowness of decision-making in organizations. In some cases, such as research, it may be possible to reduce bureaucracy by letting people write more independently in the style of an Internet forum, but it's not clear if this can work in practice or if this approach can be applied to less parallelizable tasks. I also comment on the trend for big organizations to be more risk-averse and more guarded in presenting themselves publicly.

Startups vs. big companies

When I was searching for my first software job out of college, I read a lot of career advice, including about the differences between startups vs. big companies. One of the most common statements I read was that big companies are "soul-crushing" because of the bureaucracy that's required to get anything done. Images of the movie Office Space come to mind, especially the well known scene about TPS reports.

I ultimately went to a big company (Microsoft), although I found that my team at Bing, while large, wasn't extremely bureaucratic. This was in part because my group, which improved the ranking of search results, was in many ways a research organization, where different people could explore different ideas on their own in parallel. However, there was still a reasonable amount of interpersonal coordination required at times. And I had to participate in company-wide HR bureaucracy every once in a while.

I didn't mind most of Microsoft's bureaucracy, because I took it as an opportunity to observe how a big organization works. I considered meetings a chance to learn about the interesting work that other people were doing. However, some of my coworkers were annoyed, and wished that they had fewer meetings and more uninterrupted time to get coding done.

At Microsoft, people often used the terminology of "owning" a task or project. This meant that the owner of the project was responsible for seeing that it made progress and got done. The owner would make decisions about how to proceed himself, although he was encouraged to consult with others on important questions regarding the project. This kind of ownership cuts down on bureaucracy and micromanagement because only one person needs to make most decisions and learn most of the details, while everyone else can provide shallower advice.

Of course, there's always a tension between making too many independent decisions versus too few. If you try to fully own a project, but then your boss complains about a choice you made, you'll waste a lot of time re-doing the project, compared with if you had asked the boss some questions up front. The more demanding the boss's requirements, the less you can afford to make your own decisions.

Can bureaucracy be avoided?

In Ethics into Action, Peter Singer outlines ten tips for activists that he took from the work of Henry Spira. #8 is "Avoid bureaucracy." Singer explains:

bureaucratic structures all too often divert energy into making the organization grow, rather than getting results for the cause. Then when the organization grows, it needs staff and an office. So you get a situation in which people who want to make a difference for animals (or for street kids, or for rain forests, or for whatever cause) spend 80 percent of their time raising money just to keep the organization going. Most of the time is spent ensuring that everyone in the organization gets along with one another, feels appreciated, and is not upset because he or she expected to be promoted to a more responsible position or given an office with more windows.

In contrast, Spira worked "essentially, on his own". He formed coalitions for particular campaigns, but he made final decisions. If an organization disagreed, it could choose to leave the coalition.

This approach appeals to me, and it's often how I operate as well. I tend to do what I want to do, and if other people want to help, that's welcome but not required. On the other hand, unilateral decision-making loses out on some epistemic gains that come from bringing different viewpoints together. The opposite extreme of unilateral decision-making is consensus decision-making, in which discussion continues as long as is needed for everyone to Aumann-update to a roughly shared epistemic viewpoint. Unfortunately, I find that it's almost impossible in practice to share enough information with one another to reach this level of consensus, especially since much of our knowledge is stored implicitly in System-1 neural connections that we can't easily externalize.

An interesting way to combine the speed of unilateral decision-making with the benefits of epistemic diversity is to create a platform for making individual contributions from a wide range of perspectives. A good example of this is LessWrong, where individuals make posts and comments, and then others reply and upvote or downvote. Collective wisdom is embodied in rankings and top-voted comments, but authors don't have to wait for a protracted review process that may get bottlenecked by various factors. Contributors post whenever they want on whatever they want, and commenters give feedback on whatever they want to whatever extent they have time to do so.

Academia uses a similar approach, in which individuals or small groups can publish papers without going through the bureaucracy of a big organization (except when applying for grants or interacting with their universities). However, the academic peer-review process is often slow and suboptimal, and I would guess that a LessWrong style of publishing academic papers would be more effective: publish a paper, get votes on its quality, get comments, revise the post based on feedback, etc.

Can the LessWrong model be used for nonprofit organizations? I'm not sure. A main problem is that nonprofits generally need to maintain a minimum bar of quality, and allowing employees to publish articles without oversight might tarnish the organization's reputation. That said, one could imagine organizations whose explicit structure was more like LessWrong, with donations being used to pay individuals to write their own posts. If an individual wrote enough low-quality posts for a long enough time, that individual could be let go from the organization.

Broadening this idea beyond research, one could imagine a charity that makes grants to individual people to execute their own projects. VegFund and the Animal Charity Evaluators "Animal Advocacy Research Fund" are examples of this.

That said, the fact that this grant-making approach isn't used more universally suggests that there are probably good reasons to create corpulent, Byzantine organizations in many situations, despite the attendant bureaucracy. Presumably economies of scale sometimes outweigh the costs of slowness and decision-making overhead, and in other cases, central organization may be required in order to complete massive projects. That said, scientific research is a "massive project" that is somewhat decentralized. This contrast between central planning vs. organic growth has been discussed many times, such as in "The Cathedral and the Bazaar".

Bigger is blander

The transition from small, agile startups to big, glacial companies seems somewhat inevitable and probably reflects some fundamental constraints that vary as a function of size. (If we were dealing with biological organisms, we might call these trends "allometric scaling laws".) For example, in a big company, there's a bigger cost to completely screwing things up, so you need more eyeballs on any given decision. One person's software service may be depended on by lots of other people, which prevents it from being refactored aggressively. A bigger company has more to lose if it fails than a just-started company, so it will normally tend to be more risk-averse.

Bigger organizations are also often more boring, because of risk-aversion with respect to public-relations disasters and because of the need to represent a wider constituency. One example of this trend can be seen with GiveWell. In its early days, GiveWell's blog was seen by the founders "as a place to experiment, have fun, give exaggerated versions of our thoughts, etc." As of 2016, GiveWell's blog posts tend to be extremely measured, qualified, and careful to avoid offending people. This is a natural progression that most organizations seem to go through, except maybe those whose brand is built on top of edginess.

Of course, more serious discussion of a topic can often be better than deliberately edgy discussions. However, there is also a lot lost when an organization has to present itself in a politically aware, qualified way. In particular, you lose out on the authenticity that comes from saying things exactly how you see them. For example, if I read a blog post written by a staff member of a big nonprofit, I can't necessarily tell where the author of the post actually stands on a topic, versus what the author is saying as a politically safe statement that aims to toe a compromise position among competing opinions. As a result, I take such posts with more grains of salt, because I can't tell what messages are only being said for political reasons. (The extreme example of this is actual politicians, who can't be trusted very much at all. Their statements need to be filtered through a bullshit-detecting independent analysis before I believe them.)

Personally, I tend to gravitate toward people who don't mince words and who try to "tell it like it is" (although this ideal has been somewhat tarnished and abused in 2016 by Donald Trump). For example, I would rather read a grad student's personal blog post about her research than read her published journal article (assuming the two have equal length and information content) because the blog post is more likely to say what the person actually thinks, while the journal article is more likely to hedge and say what's most expedient for getting the article published and making the article accepted by the academic community.

Sarah Constantin personifies vague authoritativeness as the Egyptian god Ra. She says: "Expressing yourself, thinking speculatively, and relating to people are shameful to the Ra-worshipping mindset, because all mental and emotional resources must be channeled into the quest for prestige." Organizations generally tend toward more Ra worship as they mature.

I don't enjoy writing in a way that's guarded and politically compromised. This is one reason I prefer to write on my own websites rather than submitting articles to be published in other fora, where editors will ask for various stylistic or content revisions that I don't agree with.