Silicon Valley Doesn't Understand Value

Abstract Artwork of a Network-like Structure

Metcalfe's Law is good for describing networked ethernet ports, absolutely useless for networked people. But Silicon Valley has relied on this law for almost 30 years to express how it thinks about the Web's value. They don't know what the fuck they're talking about.

Author's Note: This post was originally published Feb 11, 2021 on (the currently defunct) Byline.

I'm fascinated by the polling around social media companies. A lot of people don't like them but still use them every day. That's evidence of a value proposition, but it's in superposition with a trash fire. What's going on?

How do I asses the value of a social media company? The answer depends on how I ask. I can ask people how much money it would take to deactivate their account. I can ask investors to reach a consensus on the company's value. These are both interesting, but I am left unsatisfied. What I'd really like is an evaluation based on first principles.

I might try Metcalfe's Law, which influenced early Web investment funds and still does. To summarize: A network's value scales as N2, where N is the number of nodes in the network. You can see the law only concerns itself with network size. It says, “Bigger is better, and being the biggest at any cost is the best.” That makes sense once you understand that Metcalfe's Law was originally written to express value for a telecommunications company owning and operating an ethernet network. The quantum of value is a given network connection.

But, the Web is more than a collection of ethernet ports. Sure, it uses distributed computing, but our goal is to network humans. We're people, not ports! I realize that sounds a little obvious, but Silicon Valley has used this ports-not-people law for nearly 30 years to justify their approach to evaluating companies. I want to do better, so I will be more precise. When I talk about value, I mean: How valuable is the network to the people who spend time there? The quantum of value is now a moment I share with another person. Immediately, interesting implications fall out of this framing. Am I spending these moments well? It depends on the people!

Humans have a social context. And when networked, we might have a negative interaction, an interaction with negative value. Human networks are also dynamic and reactive, leading to emergent behavior. Enough negative interactions can displace and discourage the positive ones, reshaping what's possible. We see many, many, many, many, many, many, many, many, many examples of this.

And as bad as all of that looks, it's still not the full picture of the harms. Metcalfe's Law doesn't have a time dependency. But, humans have memory and culture. The time we spend with each other is an investment, a collaboration that yields structures with exponential returns on value. However, a structure can produce a large positive (Ice Bucket Challenge) or negative (Qanon) value. Dr. Elfreda Chatman's work describes how the latter can grow and persist through her “Small Worlds” theory. (Thank you to Dr. Sarah Roberts for the tip.) Current social media networks evaluate engagement, which considers these two phenomena to be equivalent. That's an outrage.

This is all opportunity cost. The opportunity cost of denied positive value plus the generated negative value. It's exponential in nature and not even contemplated by Metcalfe's Law. But, while I sense the cost is massive, I still don't have a clear model to estimate that deficit. This cost's hidden nature means that too many people end up thankful that the trash fire keeps us warm at night, without noticing how deeply we've all been wronged.

I've established that Metcalfe's Law is not useful here, but can we build a value model for human networks? Is there already scholarship around a model that includes valence and time when evaluating a social network? I see that there is a rich vocabulary for describing structures in social graphs, but assessing the quality of interaction sounds more like information theory. However, information theory alone doesn't get me what I want. It's concerned with comparisons against noise, an interaction with zero value. But what about negative value? Is there an intersectional discipline that incorporates both? Some kind of “social information theory”? I've reached the point of needing a librarian because I'm not sure where to look.

I don't have all the details worked out. But, just contemplating value through a human-centered framing gets me far enough to infer that current social media companies—with poisoned incentives around discovery and moderation—produce staggering opportunity cost. The burden is so immense, I struggle to imagine what the world would look like without it. People describing themselves as technologists have, for decades, used Metcalfe's Law because it feels like a “quantitative” way to assess value. It doesn't ask much from you. It doesn't ask you to think deeply of others or the most vulnerable. How orderly it must seem to discard essential context. But, we bear responsibility for the systems we build. To be a technologist means context is part of the deal. To be a technologist means we have an obligation; we must make that hard-to-imagine world a reality.