@bluesky

At my day job they say “Think big!”, and @bluesky is that. I can’t stop thinking about it, and bloggers can’t help thinking in public. This work can simultaneously be probably-doomed and obviously-worthwhile.

I’m aligned with where Twitter is starting this conversation. First, speaking under my AWS-employee hat: One of our standard soundbites is that people and enterprises should strive to avoid “undifferentiated heavy lifting”, i.e. wrangling software infrastructure. The basic machinery behind Twitter needs to keep track of who said what and make it available to others. Let’s go further and grant (as I would) that Twitter exhibits the essential nouns and verbs of text-centric social media: Annotated linking, following, echoing, answering, liking, muting, blocking. Which all feels to me like undifferentiated heavy lifting.

Where’s the value? · Twitter’s @bluesky hypothesis is that the real business opportunity in social media is extracting from the planetary-scale tangle a fragment stream that meets the needs of individuals sufficiently well that their attention can be monetized, and does so without being a vector for hate, bigotry, and ignorance.

It’s easy to be dubious of this claim and still convinced the project is worth pursuing. I’m both: I persist in resisting Twitter’s pleas to replace my people-I-follow-in-order feed with their “top tweets” flavor. But self-evidently I’m weird; I have the privilege and education and space and time to obsess over certain aspects of life and their truths and falsehoods, and invest in careful curation of my feed. It’s not unreasonable for people who are starved for time and less privileged to pose a simple request to Twitter (or a competitor): “Please inform me about the world quickly, in a way that enriches my life, without leading me into pathological social dysfunction.”

The second starting point I share with Twitter: Social media as currently constituted does a terrible job. The platform works better for GamerGaters than for humanists; better for Donald Trump than for the Dalai Lama; better for incels than for working mothers. So, what is to be done? The value of the quest hardly seems open to question.

What is the Internet? · It isn’t a place or a thing, it’s a lot of computers which implement protocols written down in IETF RFCs (and to a lesser degree other organizations’ publications) that reflect inspiration and evolution and are known to achieve certain effects in practice. So in the @bluesky launch thread it’s unsurprising to find belief in protocols as foundational to the project. One of the two documents they quote is Protocols, Not Platforms: A Technological Approach to Free Speech by Mike Masnick.

I’ve read Masnick carefully but I’m mostly unconvinced. He seems still to dream the foundational Internet dream that if we can just put people in touch with each other things will be better. Plus there’s market fundamentalism: Let people choose the products that most please them and this will lead to good outcomes and to products which are relatively free of dysfunction.

Which I’d like to believe but don’t. The free market is obviously a useful macroeconomic tool, but only when its freedom is carefully circumscribed. Every market needs a supporting framework of contract law and of regulation: Against fraud, misrepresentation, theft, Ponzi schemes and of course to promote safety: Aviation safety, electrical safety, chemical safety, nutritional safety, and so on. Can a protocol support the messy political mechanisms behind our imperfect but essential legal and regulatory frameworks?

The other issue that seems insufficiently considered is organized hostile action. Empirically, we observe powerful well-funded parties launching concerted efforts to bend the path of societal conversation towards Trump or away from Jesus’ Good-Samaritan lesson; or to boost Brexit. If a protocol could be effective in resisting this sort of adversary, that would be great.

I’m aware of no conclusive evidence that this is either possible or impossible. So I’d be willing to give it a shot. But I’d also broaden the focus; if there is a social-networking technology out there that can ameliorate the current dysfunction, it seems just as likely to be found in a carefully-thought-through API framework, or in some of Stephen Wolfram’s early-stage proposals (also linked in the @bluesky thread) for a level of indirection in ML-based recommendation algorithms.

Block what? · Both @jack in his @bluesky thread and his CTO following up nodded in the direction of blockchain. Any regular ongoing reader can hear my snorts of derision from anywhere on the globe. I’m not going to relitigate the argument at this point; the absence of succesful applications of the technology, all these years into it, should have removed the need. But I think the appeal to blockchain is a symptom of a powerful instinct that whatever we do to fix social media, it has to be involve extreme non-optional transparency. (Blockchain does this but unfortunately fails to be useful for other unrelated reasons.)

I share this instinct and would go further: A powerful focus on transparency and truth is an absolutely necessary (but of course not sufficient) precondition for addressing social media’s sins.

Existence proof · There is a successful online community where truth is the common currency: Wikipedia. Yes, it’s flawed in multiple important ways, but its insistence on publishing only what’s backed by evidence, and its rough consensus on what evidence is acceptable, is at the end of the day heartwarming. Its great flaw is not inaccuracy but inattention; the community of editors is generally insufficiently diverse and specifically mostly-male, and this reflects into Wikipedia’s pages.

Wikipedia’s contract with the world is simple: Any material that is not verifiable will eventually be removed. So obviously the meaning of “verifiable” matters and is thus a source of controversy. Healthy, reasonable, sensible controversy. Which we observe to have an imperfect but highly usable result, relatively free of the currient pathologies of other civic discourse.

Am I saying that the @bluesky solution comes down to “remove anything not verifiable”? No, but I’d be inclined to suggest that “verifiable vs not” should be an important input to the @bluesky algorithms and protocols.

Testing · Suppose the @bluesky team ends up looking at a few big ideas — let’s ignore whether they’re protocols or algorithms or ML models or some other sort of thing. My first question is “how do you tell if they work?” At AWS, something relatively simple, like the performance of a throttling algorithm under stress, can be difficult to test without, you know, putting it under stress.

We’ve developed incremental nondestructive imperfect techniques for evaluating this kind of thing, and you’d need something like that to make useful decisions about @bluesky ideas. You’d need to run them at scale in the context of an automated harness, and you’d need to do it more or less all the time, forever.

The construction of this infrastructure is, I believe, apt to be one of the hardest parts of @bluesky.

Will it work? · I doubt it. Social-media dysfunction seems closely related to aspects of human nature at the individual and societal levels. A solution might end up being more or less equivalent in difficulty to World Peace or Curing Cancer.

Which is obviously not a reason to not try.

---

---

Contributions