WATCH ⏯️ Highlights from last week's evidence session with researchers on the #OnlineSafetyBill 🧵
Witness @LauraEdelson2 @nyutandon explains to the Committee how 'edge effects' and user interactions shape content algorithms 📱
1/7
Witness @LauraEdelson2 @nyutandon explains to the Committee how 'edge effects' and user interactions shape content algorithms 📱
1/7
⏯️ Witness @LauraEdelson2 tells Baroness Kidron that without seeing the data behind it, research from social media platforms ‘is not science’ 🔎
2/7
2/7
⏯️ Chair @DamianCollins: "the more vulnerable you are, the more likely you are to see harmful content [...] Would that be a correct assessment?"
Witness @LauraEdelson2 @nyutandon: "Frankly yes, this gets to the heart of interest-based content promotion"
3/7
Witness @LauraEdelson2 @nyutandon: "Frankly yes, this gets to the heart of interest-based content promotion"
3/7
⏯️ Witness @gchaslot @aiTransparency: "Studying the @YouTube algorithm, it doesn't try to find the perfect content for you, it tries to find the rabbit-hole that's the closest to your interests" 🐇🕳️
4/7
4/7
⏯️ @Q66Suzi asks "How long will it take to take content down?"
Witness @gchaslot: "It doesn't matter how long it stays on the platform, what matters is how much the algorithm is amplifying"
5/7
Witness @gchaslot: "It doesn't matter how long it stays on the platform, what matters is how much the algorithm is amplifying"
5/7
⏯️ In response to Lord Gilbert of Panteg @Stephen17262836, witness Renée DiResta @noUpside @stanfordio explains why 'reduce' is a better tool than 'remove' for moderation of potentially harmful content.
6/7
6/7
⏯️ Witness @LauraEdelson2 @nyutandon: "I would strongly encourage you to include advertising in what is under oversight. There just isn't an easy distinction to be made" 💸
7/7
7/7
• • •
Missing some Tweet in this thread? You can try to
force a refresh
