Social media spam as a case examine
Disclaimer: the examples on this put up are for illustrative functions and will not be commentary on any particular content material coverage at any particular firm. All views expressed on this article are mine and don’t replicate my employer.
Why is there any spam on social media? Nobody except for the spammers themselves enjoys clickbait scams or phishing makes an attempt. We now have many years of coaching information to feed machine studying classifiers. So why does spam on each main tech platform really feel inevitable? In any case these years, why do bot farms nonetheless exist?
The reply, in brief, is that it’s actually onerous to struggle spam at scale, and exponentially tougher to take action with out harming real customers and advertisers. On this put up, we’ll use precision and recall as a framework for understanding the spam downside. We’ll see that eradicating 100% of spam is impractical, and that there’s some “equilibrium” spam prevalence primarily based on finance, laws, and consumer sentiment.
Think about we’re launching a competitor to TikTok and Instagram. (Overlook that they’ve 1.1 billion and 2 billion month-to-month lively customers, respectively; we’re feeling bold!) Our key differentiator on this tight market is that we assure customers may have solely the best high quality of movies: completely no “get wealthy fast” schemes, blatant reposts of present content material, URLs that infect your laptop with malware, and so on.
Try 1: Human Evaluation
To realize this high quality assure, we’ve employed a staggering 1,000 reviewers to audit each add earlier than it’s allowed on the platform. Some issues simply want a human contact, we argue: video spam is simply too advanced and context-dependent to depend on automated logic. A video that urges customers to click on on a URL could possibly be a malicious phishing try or a benign fundraiser for Alzheimer’s analysis, for instance — the stakes are too excessive to…
