Why Nostr? What is Njump?
2024-03-17 02:54:29
in reply to

Rabble on Nostr: I’m not sure about algoeithims but algorithms simply aren’t a thing we can avoid ...

I’m not sure about algoeithims but algorithms simply aren’t a thing we can avoid in software. Fundamentally software is a series of mathematical equations organized together in algorithms. It is kind if like saying you want an ocean without water. There are oceans of other liquids on other planets, but here on our planet, with computers as they are built today, all software is algorithms.

When it comes to social media and the way out software decides what to show us, every possible way choosing data and showing it to a user is an algorithm. It might be just show the posts from people you’ve chosen, either in the order in which they publish them or in which the client receives them, those are algorithms. That choice has a bias and rewards some behavior while punishing others.

Take for example who posts a LOT of short posts. If we use the most frequent posters get more exposure, then we’re focusing on building an audience of direct follows, getting your posts shared, and posting a LOT. We could have an algorithm which focuses on showing what’s trending in your network, or one which prioritizes showing you posts with lots of comments, or a feed that shows you content which has the most reports.

There is no escape from using algorithms to decide what content you will see.

What we need is a system where users have agency, where you can choose your algorithms, see the trade offs in each, and the way to pick what works for you. The algorithm isn’t the problem, it’s that is a monoculture that’s picked by platforms. It flattens culture and makes us vulnerable.

Facebook actually had researchers try presenting users with content / social graph/ engagement aware algorithm vs chronological feed. What happened was a chronological free showed less news, less diverse view points, and lots more viral shared content. It had no change in people’s political views at all. The biggest take away is that the cohort of users given the chronological timeline used it less and switched their time to social media apps that did give them a more customized timeline algorithm.

https://www.gsb.stanford.edu/insights/surprising-discovery-about-facebooks-role-driving-polarization

We can’t escape the algorithm. First because that’s how all software works, but also because users prefer something other than the chronological timeline we think we want.

Over at Bluesky they’ve got 40,000 algorithms to choose from in a way that’s easy to discover, use, and make. Nostr has a few dozen to choose from because for the most part we combine the algorithms with the clients. Only a few clients use the DVM feed functionality that lets us decouple the algorithms from the client.

We should embrace and support user agency to choose and create algorithms that they can control. We need DVM feeds to be shown in clients, documented, integrated, and to be fast. Relays need to be able to support requests of a list of note ids not just full note content so we can make the system fast and low bandwidth.

We can do this. You can choose an algorithm which pulls from the lists your friends manually curate, or from some advanced LLM predicting your interest. If we have a hundred thousand algorithms then the power of each one diminishes. Maybe you want to have a feed that helps you find the shit posters, while I want to hide the flame wars. Both are fine.

Let’s not reject algorithms, let’s admit that they’re baked in the fundamentals of how the technology works. Let’s take power over them instead of the large platforms using them to control our public sphere.
Author Public Key
npub1wmr34t36fy03m8hvgl96zl3znndyzyaqhwmwdtshwmtkg03fetaqhjg240