Thank you for your thoughtful input.
I thoroughly understand your argument. My team and I have been thinking deeply about these issues a long time in respect to Medium, other systems we’ve been a part of developing, and the internet as a whole. I say this not to imply we know better or that we are necessarily right. Just that we designed this system carefully with the same concerns you have in mind.
Safe to say, we are all speculating about what will happen. (One thing experience in this realm leads to is humility about one’s ability to predict human behavior in complex systems.)
Your data point is interesting, but to my mind it just underscores why a binary system (i.e., one user = [maximum and minimum of] one vote) will not suffice. You’re not suggesting that. You’re suggesting time-spent as the more accurate measure of value.
As long-time proponents of the time-spent measurement as a useful proxy for quality, it might be curious why we are not just using that for this new purpose.
What we’ve learned over the last few years of trying to figure out how to measure quality — looking at lists of posts ranked by time-spent, versus views, versus recommends, and other permutations and combinations is that…it’s complicated. Roughly speaking, our subjective conclusions are that time-spent is better than views but, in its own, not as good as recommends. And that any single metric has pros and cons. A major caveat to that we’ve measured time-spent and views across all users, while recommends are only from those logged in.
Another caveat: We’ve never paid out based on a metric, and I believe as you do that MONEY CHANGES EVERYTHING.
That said, here’s one thing that’s unambiguous from our analysis: Optimizing for time-spent rewards longer posts.
Is this good?
It would be if our goal is, as you put it, to be “the Web’s last, best shot at a home for serious, long-form writing.” But that’s not really our goal. Serious, yes. Long-form? Maybe.
We are fans of long-form, but not only long-form. Our real goal is to support the creation and distribution of content (writing, and other forms) that deepens understanding and spreads good ideas.
As far as I’m concerned, doing those things in less time is better. Conciseness is a virtue. So building a system that optimized for length seems unideal. Consider, like anything else, time-spent would be gamed. And that, for instance, while it takes a lot longer to check a fact, for a writer, it takes no longer to read a true fact than a false one.
If you’re reading for enjoyment/entertainment, longer might be better. We’ve all had that I-don’t-want-this-novel-to-end feeling. But how often do you have the I-don’t-want-this-article-to-end feeling — even when it’s great?
That’s my argument against reader minutes as the metric. However, let me clarify something else that seems to have been largely missed: There probably is no the metric. In all our documentation, we’ve said that payment will be based on engagement, not that it will be based (solely) on claps.
A joke about claps makes a good headline for those who are under pressure to dash off click-driving content as quickly as possible, so that impression has been (ironically), widely spread. Also, admittedly, we’ve put an emphasis on claps in our own explanation as the primary metric, because we want members to know they have agency in where their money goes. That it can be a conscious choice — not the result of accidental behavior — is very powerful and unique. It’s what changes it most fundamentally from the current paradigm.
We’ve also heard from the beginning of our membership initiative that people want to support authors, and they want more control. (And, yes, we could have designed in even more control, but we also have m to balance that with simplicity and low cognitive load. Our challenge became: Can we design something that feels natural and becomes automatic — with familiarity — but is still explicit?)
This doesn’t allay your valid concern about manipulation.
One correction, though: Authors can see how many claps came from each person, but that information is not public. The thinking here is that it’s useful to see who your super-fans are, and it’s nice to let an author know you really appreciated their work. But it’s potentially distorting to show that publicly.
However, your point about emotionally charged (and, I would add, belief-reinforcing) posts potentially being rewarded disproportionally still holds. The good news is, you, as a writer would not necessarily be effected by this, because it is not a global reward system. If your readers are not rewarding cheap content, it doesn’t really matter if others are, because the budget for each member is doled out relative to their own behavior.
This reflects our default philosophy of erring on the side of trusting our users and building the system we would want to use. I’m personally optimistic that more important work gets rewarded if we put that power in the hands of our members than if we try to design the ultimate implicit system.
The other good news I would offer is this: None of this is set in stone. I’m defending the thinking behind the system we’ve deployed, but we are readily willing to change it in the light of data. We will, of course, be analyzing the results carefully and refining it over time.
We will also share what we are learning and how we are evolving things. And we always welcome your (everyone’s — especially writers and member’s) feedback and ideas on what could make Medium better.
Considering multiple perspectives and ideas is, after all, core to our ethos.