Nice post, Arthur.
Added cycle_dao to my followees (why aren’t they just called follows?). Great work on that blog post. The two things that most convinced me to follow were your “no” votes on:
- NNS Proposal: People Parties
- NNS Proposal: Change Dissolve Delay Bonus and Age Bonus Parameters
(By the way, might be nice if your posts included the position you took in the headline)
Defaults are powerful things.
For now, I believe the DFINITY Foundation has wielded their influence well. For example, their frequent strategy of following the community essentially means that uninformed voters just support the people that are informing themselves. For that reason I still follow them, and think they’re a good default follow.
I want to support adding additional defaults, but two things concern me:
- Power superlinearity
The UX concern is simple: getting up and running with NNS is already a very involved, ambiguous, and worry-inducing process for new users. Ambiguity kills user confidence.
The UX we want to offer is this:
- User learns about ICP, buys some.
- User learns they should be staking their ICP for voting control and for rewards, invests time into learning how the internet identity and the nns works
- In the NNS, user learns how to navigate and stakes their ICP to starts earning rewards
- User learns that they’re following Internet Computer Association (ICA) or DFINITY Foundation (DF) by default and so they go hunting to see who they would prefer to follow and thereby invest in the community and governance conversations.
This creates the incremental trust-building user flow that we’ve all experienced, it’s what got us here. The learning curve for the internet computer is steep, this Learn → Invest → Learn loop makes it a bit easier. Each step in the above is a completion of the Learn → Invest loop.
I don’t see much benefit at the moment to moving the selection of followees from the fourth learn → invest loop back to the third.
I can see a UX cost, which is that as more neurons are available by default to choose from, users could feel like it’s too much work to finish up the staking step.
2. Power superlinearity
Even solving for UX, changing the default away from Internet Computer Association still leaves us with a problem: power superlinearity. Let’s tell a story, but now I’m adding a malicious neuron into this proposal named “DFINITY Team”, which is run by me and just trying to accrue power:
- User goes to NNS to stake ICP, they see they need to select a neuron to follow
- User is presented with cycle_dao, icdevs, DFINITY Team (my malicious neuron), ICA, and DF. They might select semi randomly, but they’ll likely favoring DF, DFINITY Team, or ICA due to name recognition. Otherwise, they’ll delay selecting a neuron due to their concern about choosing poorly.
- My malicious neuron, DFINITY Team, steadily gets more voting power from non-vigilant users which I can do whatever I want with, because the users aren’t paying attention. For example, given sufficient time perhaps I could introduce a motion to kick DF, ICA, cycle_dao, or icdevs out of the defaults.
This is a concerningly plausible story that just comes from the power of defaults and name recognition. Obviously, this isn’t what you’re doing. But it’s possible for someone to do if we stick with the current flow of needing to propose neurons as default follow-target neurons.
I think the existing user flow is pretty good as it is. It makes sense that neurons that haven’t intentionally selected a followee should just follow the community’s vote. However, there’s still an opportunity here:
What we could do is make it easier for a user to discover the neurons they would prefer to follow. This would make completing that 4th learn → invest loop easier.
I just found out about cycle_dao through this post, excited to have found it! What other methods might there be to set up a credibly neutral mechanism for exposing new ICP stakers to neurons they might like to follow?
We have a big opportunity here to design the context within which neuron followership is evaluated and earned, including the process by which neurons are selected as defaults.
Inside that context, it might be much easier for people like me to find cycle_dao in the future