How to Kill Fake News with Robots
Building a better system of ad-blocking
Last night, I attended a Meetup discussing the Future of Ad Blocking.
TL:DR; It’s hard! But people are trying.
One of the promising techniques is to control the format. (This is a start!)
The purpose of this post however, is to propose the creation of an opt-in robot-hive-mind which acts as a kind of automatic, invisible police-force for bullshit.
Done right, this could fix crappy ads and “Fake News” simultaneously.
Three core principles:
- Blocking is good. Tracking is better.
- Algorithmic power scales with Metcalfe’s Law. (No relation.)
- Machine learning is our friend, and can fight proxy wars on our behalf!
Got it? Next up is system architecture.
Picture a browser extension that:
- After a page loads, back-tracks all pixels to identify the source root domain structures, encrypts the path+source data, and sends it to M.O.M*. (Massive Online Megaserverwithmachinelearningcapabilities)
- M.O.M ‘Sandboxes’ those domains, cross-checks collective user-data for similar pixel>domain routes, and uses it to create a real-time, global database of all domains being advertised, including how, where, and whom is doing the advertising on their behalf.
- M.O.M. then runs an ‘authority-check’ based on collective traffic for that domain structure, and known behavioral profiles of spammy / phishy / misleading / generally-not-awesome behavior, and creates a ‘naughty list’ of domain structures in this category.
- M.O.M of course, updates the browser extension with the ‘naughty list’.
- Browser extension now auto-blocks all ads pointing to the naughty list (even if they’re brand new, different ads) and perhaps, occasionally sends a list of the worst apples to the FTC and its foreign counterparts.
This allows a kind of pre-emptive strike against new ads pointing to well- known bad-actors… it also has the ability to add a ‘bullshitometer’ to the right-click, which could inform the user if say, an article was paid to be surfaced to them, or if the domain structure has only existed for a week.
Currently most Ad Blockers don’t track either user behavior, or the ads they block. (This is a good thing, and speeds up your internet experience.)
That said, this kind of approach would require a tiny amount of our collective fire-power to run the checks… which could mean allowing things like anonymized click-through data (informing M.O.M’s Bullshitometer of the CTR-to-bounce-rate ratio) or using post-load CPU capacity to capture, encrypt, and send data back to M.O.M. — even if the ad is never surfaced. (There’s ways to do this that don’t slow down your interwebs.)
There’s a few folks who have this kind of fire-power today. One of them is eyeo (owners of AdBlocker Plus) — the others would be Mozilla, The Googs, and perhaps Apple… you could also try buying a bunch of botnets, but I hear that’s frowned upon.
As for monetization, there’s certainly plenty — but I haven’t figured out an ethical version, just yet.
…Perhaps you could find one that works?