Read on Twitter
The Recommendation Algorithm is broken, here’s how to fix it.
A thread...
The Recommendation Algorithm must be reconfigured to bring us together instead of driving us apart. As a software engineer of 30 years I believe this is actually achievable.
A thread...
The Recommendation Algorithm must be reconfigured to bring us together instead of driving us apart. As a software engineer of 30 years I believe this is actually achievable.
The Recommendation Algorithm is currently dumbly configured with the single goal of maximum engagement in the pursuit of advertising revenue.
To keep us engaged The Algorithm attempts to show us only things we’re interested in, and, more importantly, only things we agree with.
To keep us engaged The Algorithm attempts to show us only things we’re interested in, and, more importantly, only things we agree with.
When we choose to like or share a news report, The Recommendation Algorithm immediately promotes, to the top of a list of things to show us next, the most popular news reports among people that also liked or shared the news report we shared or who are deemed to be most like us.
It should come as no surprise that there are tragic consequences arising from the choice to marshal the flow of information reaching us though the global network, based on the whims of Zuckerberg’s creepy vision for how to rate “who’s hot and who’s not” on his college campus.
There is, however, a potential solution lurking in the data, if only we push the process beyond the base popularity contest and instead search for consensus.
We’ve all heard how promoting the most popular items leads to contagious lies flying around the global network six times faster than more nuanced truths.
While engagement and it’s brutal henchman, popularity continue to run the show, bullshit will always fly faster than fact.
While engagement and it’s brutal henchman, popularity continue to run the show, bullshit will always fly faster than fact.
Rather than use popularity as it’s primary metric, The Recommendation Algorithm must be reconfigured to prioritise consensus; to identify not the reports popular among those with matching profiles but the reports on that topic shared by both those like you and those unlike you.
The Recommendation Algorithm must be reconfigured to amplify what we share, to highlight where we agree, to show us what we have in common, to bring us together, to promote unity instead of manufacturing and magnifying division.
A source of hope: Reconfiguring The Recommendation Algorithm toward algorithmic accentuation and amplification of consensus, of where our views overlap, would compound over time, just as we have seen with the algorithmic accentuation and amplification of our divisions.
