AL #015: Threads is a Masterpiece in Mechanism Design
Web3 protocols are modeling humans wrong
A new social media app launched two weeks ago by following a familiar playbook:
Use a trendy domain name associated with finance or decentralized networks (they used .net)
Start with a centralized implementation but promise to decentralize soon and to deliver full portability of the underlying social graph
Mint a POAP for each user and display it visibly on their profile
The app came from a company that previously unsuccessfully tried to create an L1 and build several meta-verse apps so is well-versed in the crypto space.
I'm talking about Threads by Meta of course.
While it's clear that Threads has adopted a lot from the web3 community, I think the web3 community can learn even more from Threads.
The Mechanism Design of Threads
While mechanism design in web3 is optimizing for utilons (assuming perfectly rational and profit seeking agents), Threads is optimizing for hedons (assuming human agents).
Meta understand that dopamine offers stronger reinforcement than virtual money.
Mark has already crossed the 100M user mark with Threads by doing the following (among other things):
Showing your Threads joining id in your Instagram profile as a link to your Threads account. This serves the dual purpose of encouraging people to join early to claim their id and advertising Threads profiles on popular Instagram accounts
Limiting consumption to a single algorithmic feed. Meta has decided early on that users who “curate” their own experience cannot be manipulated into taking certain actions that are beneficial for the platform. Instead every user is given a single feed to work with and the only custom controls they have are Follow and Mute (for now).
Making the algorithm friendly for new creators by leaning towards exploration. In particular, the Threads home feed is far inferior to the Twitter “For you” tab. There is good content on threads but the algorithm is deprioritizing it in favor of randomly showing you both popular and unpopular posts. As a result, new users are getting far more followers than they would be able to on a more discerning platform and they love it.
Potentially using AI to filter content (speculative). Users have noted that Threads content is more “positive” and this is something that may have been achieved using a sentiment scoring engine. It’s highly likely that AI is being used to add additional attributes to the scoring function in the algorithm, making certain types of content more likely to appear.
Since Threads has no active monetization, profit focused mechanism design has little to say about why users are flocking to the network.
Instead, we can lean on the Hook Model popularized by Nir Eyal.
The Hook Model argues that the most successful consumer apps drive users through a funnel of 1) Trigger 2) Action 3) Variable Reward 4) Investment.
Each step in the Hook Model violates several principles in mechanism design.
Step 1. Users need triggers to take actions
Humans don't have full visibility of our action spaces (technical term for “all the things we could do”).
We tend to react to prompts from the outside world instead.
Users need a trigger that makes them aware of what action they could be taking.
Step 2. Users are compelled to take actions
It’s not enough for an action to be “rational” for a given user, they need to be motivated to take it.
In many cases an impulsive or intuitive argument can be more effective than explaining why a certain action is utility maximizing for a given user.
Step 3. Users are addicted to variable rewards
As various forms of gambling (from casinos to “Gacha” type drops in games) have illustrated, humans enjoy variable rewards.
Mechanism design literature instead argues that agents seek to minimize uncertainty.
Both are true but there is a difference in time scale.
On shorter time scales, people enjoy variable rewards.
During longer time scales (e.g., career choices), people tend to be more risk-averse.
This applies to product usage too.
Step 4. Users are loyal to products and communities that gave them positive emotions
Research behind vampire attacks on crypto protocols suggests that users will abandon one platform readily for one that offers incrementally more rewards.
In practice, web3 communities exhibit evidence of loyalty/tribalism and the Hook Model's word for this is “investment”.
The Hook Model predicts that users continue to invest time & resources into products that delivered them positive variable rewards.
While Threads has been great at Step 1-3, it's Step 4 where they will really be tested.
Retention.
Meta is one of the few major social media companies that had to rename itself because of how irrelevant its original product had become.
While Instagram has done better, the success of Threads will likely rest on their ability to retain users.
Can they move from the current engagement-driven feed to one that's more sustainable and enjoyable to long term tenants of the network?
Join me and lets find out:
If you're reading this on your phone, you can find me there at https://www.threads.net/@peteriserins.