Social media, algorithms, and NY
Arran Hunt
Partner at McVeagh Fleming. Immigration and Internet law expert. Superdiversity Institute Top 100 Board Ready Director.
Late last week I was quoted in an excellent piece by NewsHub on Deepfakes https://www.newshub.co.nz/home/politics/2024/06/australia-criminalises-ai-nudes-should-nz-do-the-same.html. While that is a hugely important topic, it isn't one I want to cover today. However, this article does relate to it, or at least had me considering it over the weekend. I promise that we'll get to New York soon, but first I need to explain why I was thinking about this.
I have a large following on Facebook, with about 3000 friends and 4200 followers. It's a platform I use mostly for work, but with family using it to stark in contact as well. I've had a look at my last four posts that weren't shares (as shares don't show this information):
11 May post of aurora australis from my driveway, 7037 impressions, 6801 post reach, 200 engagements
29 May post about confirmation from Immigration New Zealand to an error we found in their rules, 6777 impressions, 6685 post reach, 181 engagement
4 June post linking to an NZHerald article about migrants, 536 impressions, 527 reach, 123 engagement
8 June post linking to the NewsHub article, too soon or low to get impressions and reach, 4 engagements.
We could say that the first two were from longer ago, but most impressions occur in the first 24-48 hours. We also can't look at the post type, as there are two posts about immigration, one with great impressions, and one with poor. Neither can we say that engagement was too low on the most recent posts, as 123 engagement from 536 impressions is substantially higher than the previous two. By all reasoning, the NZHerald article on migrants should have been encouraging Meta to show it to more people. Yet it wasn't. The most recent post was barely being shown to anyone.
As a little completely unscientific test, I've just posted a short post about immigration, at 815am on a Monday morning, so not really peak timing. I'll post the statistics at the end.
领英推荐
Algorithms
Of course, this is likely down to the ongoing conversation about social media companies and the news media. The more we see these posts, the greater the argument for the news media for pay from the social media companies. So, anecdotally, these posts don't get shown as much. Such posts also draw the user away from the social media platform. For several reasons it is better for social media platforms to not show these posts.
There is also possibly a factor in what the articles are about. Immigration news isn't something that directly relates to Meta. While they'd rather you don't leave Facebook, the article itself is distant from the platform. However, deepfakes, and their use to cause harm, and Meta's goodwill is more likely to be impacted. With the intelligence of the AI algorithms being used, I wouldn't be surprised if there was some scoring system used to determine which articles are better to link to, and which might be detrimental to their business. They have a personal interest in not showing such posts, or at least to show them much less frequently.
We're using these services that chose to manipulate which of our posts it shows and which is doesn't. While there is no freedom of speech argument, as these are private companies, they are still choosing as to which bits of our speech they want to publish more, and which to suppress. This is, of course, within the terms we agreed with. However, I really didn't realise how much difference it would make. Facebook used to appear to show me the posts of all my friends, but now it is manicured to both those I interact with the most, as well as sponsored posts that they think I might like. It really is all about retention on the platform.
New York
This leads us to New York. The state government there has just voted in support of new legislation controlling how social media interacts with those under 18. Most notably, it seeks to ban what is sees as "addictive" recommendation algorithms. The "Stop Addictive Feeds Exploitation (SAFE) for Kids Act will prevent social media platforms from delivering content based on algorithms. Instead, it will just show posts in a reverse chronological order, news to oldest. it will also only show posts from people they are friends with or follow (depending on the platform). The social media company will lose the right as to which posts to feature and endorse. It would be taking social media back to the original days when we would see new posts from friends, not six month old posts from companies gaming the algorithms. It would also ban sending notifications between midnight and 6am. Parental consent can be given to bring a child outside of the new rules.
The new law is expected to be signed by the Governor this Friday.
I'd expect that we'll see similar legislation popping up across the USA, and perhaps Europe will follow. While I don't expect any positive changes for the use of those 18 or older, it would be nice to go back to an algorithm that actually showed me all of my friend's posts.
As to my post from 16 minutes ago, I have 39 engagements already. Clearly, Meta likes my immigration posts that keep people on their platform.
UPDATE - One hour in and the post has 479 impressions and 94 engagements.