The book Liars and Outliers: Enabling the Trust that Society Needs to Thrive provides a framework to answer the question, “Why do people trust each other and cooperate?”
I read this book with an eye towards improving my understanding of how people filter information, which is relevant to the focus of this blog and my recent interest in improving the trustworthiness and quality of crowd-sourced product information. I also knew of and respected the author, security expert Bruce Schneier, who is a source for parts of my password management series.
Filtering information effectively requires trusting your information sources as well as the people who recommend these information sources. If it were fully understood why people trust each other and cooperate, that might guide the development of much more effective and automatic systems to make online information sharing more trustworthy and relevant.
So what did I think of Schneier’s book?
Summary of Liars and Outliers (with Online Example)
Liars and Outliers frames trust and cooperation as competing interests (self, relational, competing groups, competing morals) resolved by four societal pressures (moral, reputational, institutional, and security systems). Faced with a decision, an individual decides to cooperate with or defect from a given interest based on these four societal pressures.
I’ll illustrate the conceptual framework with an example of an online system that I believe does a particularly good job at fostering the exchange of trustworthy, high quality information: Stack Exchange.
Stack Exchange (most famous for Stack Overflow) is a question and answer site that uses a number of methods to optimize the quality of both questions and answers. It has achieved a level of quality and trust that is tops among Q&A sites and is far above the average online site.
Stack Exchange actively discourages asking questions whose answers are unlikely to be objective and useful. This includes all subjective questions. However, those who ask questions often want answers to questions with a subjective element.
So here we have an example of a “societal dilemma” where an individual’s self-interest competes with an organization’s self-interest. The individual wants an answer to a subjective question. The organization doesn’t want site quality to decline from the corrosive influence of subjective questions.
Using Liars and Outliers terminology, here’s the societal dilemma:
Societal Dilemma: Subjective questions
Interest: Stack Exchange wants to uphold high quality.
Norm: Cooperate by posting only objective questions that have objective answers (avoid forum-style discussions which inevitably cause quality to degrade over time).
Competing Interest: Individuals want to get great answer(s) to a subjective question via Stack Exchange.
Corresponding Defection: Post a subjective or broadly scoped question.
To encourage people to act in the group interest, Stack Exchange implements the following societal pressures:
Societal Pressures: Stack Overflow on subjective questions
Moral: Uses many methods to educate people and encourages them to feel good about producing a high quality Q&A database. Examples: blog posts, meta boards (answer questions on how to use site), quality metrics are shared, people are required to “sign up” to follow a new Stack Exchange community.
Reputational: Makes reputation explicit with scores. Tweaks (“turns the knobs” of) reputation system to reward good quality and punish bad quality. Online reputation is somewhat tied to real life reputation in the form of job opportunities.
Institutional: Lists policies and FAQs for each stack (and Stack Exchange as a whole). Questions can receive negative comments as a form of shame. Questions can be “closed” when not following the norm.
Security: Algorithmically detects spam. Automatically delays new posts from appearing when appropriate (single user posting too frequently, rapid back and forth discussions, etc.).
So far as I can tell, Stack Exchange relies primarily upon moral and reputational pressure. This is significant and interesting as research suggests that these two pressures rarely scale well, yet Stack Exchange now has millions of users. By relying less on institutional and security pressures and more on moral and reputational pressures with its explicit reputation system, it has become an uncommonly effective online institution (for more about online services as institutions, see pages 203-204 of Liars and Outliers).
If you’re intrigued by the above example but feel a bit confused by the model and the terminology, that’s okay. The first chapter of the book lays out the model and terminology, while the first half of the book describes the model in great detail.
If you’re able to correctly identify all of the different interests and pressures and their strengths, then according to this model it should be a simple matter to predict whether a given individual will defect and cooperate for any single decision. More broadly, you may be able to craft better policy for your organization, family, or online community that results in increased cooperation.
Strengths and Weaknesses of Liars and Outliers
I love this book’s elegant conceptual framework and it made sense to me both intuitively and intellectually, especially given the wide variety of cited research. But how do you get from this conceptual framework to practical results?
That’s what I kept wondering as I plowed through the first half of the book, and kept asking myself, “so how do I accurately identify all of the different interests and pressures and their strengths?” To be fair, Schneier did not set out to answer this question when he wrote this book. But it’s difficult for me to fully embrace the concepts if it’s impossible to actually test them in the real world. Testing Newton’s law of universal gravitation is simple. Testing this conceptual framework is not.
Perhaps my expectations are too high. Perhaps any attempt to bring rigor to the social sciences results in proposed frameworks that are difficult to test.
The other issue was that I found this book tough to read. Schneier goes into great detail, perhaps excessive detail at some points, and I was always afraid I’d miss some key insight. So I tried to read every page. The result is that I kept putting the book down. It took months for me to complete it. In the end, I skimmed the last few chapters. Perhaps rapidly skimming the book and revisiting the sections of most interest would have worked better.
If you choose to skim or speed read, be sure to read the first chapter carefully. Thoroughly understand Figure 1 on page 12. With that understanding, you can dive into any other part of the book in any order you like.
Conclusion
Liars and Outliers presents a conceptual framework which elegantly pulls together disparate research from a broad range of disciplines into a cohesive whole. That alone makes it a very good book. Even better is the enormous amount of meticulously referenced research.
But despite my strong interest in the subject matter, I found I really had to push myself to finish reading the book. The book is not structured and written in a way that makes for an easy read. For that reason, it doesn’t quite belong in the same league as The Selfish Gene, or Guns, Germs, and Steel. These are books I’ve read which served a similar function—comprehensively bringing together the latest research from a broad array of disciplines to explain complex, widely scoped phenomenon that are not fully understood.
Nevertheless, this book is a “must read” for academically inclined individuals who want an introduction to trust and cooperation. If you tend to get a lot out of books like The Selfish Gene and Guns, Germs, and Steel, you’ll find this book just as valuable in helping you construct mental models for how the world works. Just be warned that it’s not as easy to read.
Disclosure: I was among a group of people who were each sent a copy of Liars & Outliers signed by Bruce Schneier at a reduced price. In return, I promised to review it. I believe I wrote this review the same way I would have written it without having received the discounted, signed, copy.