Hacker News Reader

Comment by alex_g | Hacker News Reader

This is a weird take.

You get lots of messages from Nigerian scammers, but the solution was not to prevent people from writing freeform emails. The solution was to build powerful spam detection algorithms, make it easy for people to classify emails to help strengthen the training set, and the problem is basically solved.

There's no easy answer to content moderation. There's no one size fits all solution, nor is there is some weird hack that's going to fix it. It's a part of your product. If you treat it as such, you're better off.

If you treat it as a separate problem that just needs money thrown at it or duct tape wrapped around it, you're never going to stop throwing money and tape at it.

Everyone wants an easy way out. You need everyone on your team in the room brainstorming solutions.

View on HN

The problem isn't solved, just surpassed, given that Nigerian spammers continue to make money.

Fair, though I'm curious if the people falling for them are using email providers with quality spam filters. I'd guess it's a much older crowd that's more likely using an archaic email provider with no incentive to improve spam filters.

> I'm curious if the people falling for them are using email providers with quality spam filters.

I'm curious why the people falling for them need a spam filter to recognize them as scams. I still see an occasional one slip through my email provider's spam filter, and I've never had any problem figuring out that they were scams.

Because not everyone has competent internet skills.

I don't think it's internet skills per se, given that scams exist since before the internet.

People default to trusting other people

Part of internet skills is increasing your prior that someone is lying to you.

The threat models from 419 (Nigerian / advance-fee fraud) spam are:

- Individuals are subject to a fraud risk.

- Trust in the medium as a whole is decreased.

- Risk of false-positive spam flagging increases, decreasing legitimate message transmission reliability.

- Fraud will continue so long as costs of operation are lower than rewards.

Economic spam, in a word, follows an economic model, and is subject to economic attacks. The notion of some form of "email postage" is one ultimate form (thus far both impractical and widely rejected), but other "increasing the costs of business" have had ... significant if not perfect ... results in limiting the activity.

The calculus for political propaganda is vastly different, with the rewards of even a marginal success being tremendous, and the resources of antagonists being prodigious. It is very much like war, that most costly of economic activities.

It is war.

It is information warfare, rather than physical warfare. The weapons are messages, messengers, channels, informants, and audiences, rather than missiles and knives and fire and water and chariots (or tanks or helicopters or bombers or ...).

But Sun Tzu wrote of both types 2,500 years ago.

Unfortunately, the fundamental architecture and premises of a major contingent of participants fundamentally enables the attackes presently being conducted. (This is of course something of a tautology: if a terrain is fundamentally favourable to a specific mode of attack, that attack will tend to be employed, successfully, on that terrain.)

There are multiple ways to address problems with objectionable content. One is not to have it, which is what the author proposes.

Spam detection works well to eliminate email you don't want to see but nobody seems to have figured out how to apply it to social media content. Part of the problem (I guess) is that most of us agree what spam is whereas objectionable content is very much in the eye of the beholder.

I like the idea of an objectionable content filter that's personalized like a spam filter. People don't always agree on what spam is, and they can change their minds at any time.

Reply on news.ycombinator.com