How Twitter could solve its spam problem, for good
Jun 24, 2011
4 minute read

Twitter spam is a frequent occurrence, and I typically ignore it, but not today. This morning I received this message from an account that has been spamming people for several days. I did my duty and reported the account for spam and…nothing.

I did a little research and discovered that the director of Twitter’s Trust and Safety team is @delbius (Del Harvey), so I asked her, “Can Twitter get more aggressive against spam? e.g., autosuspend reported accounts that frequently mention people they don’t follow?” She was kind enough to reply that they do, and that they’re working on improving the rate of detection.

I reloaded the offending account, and sure enough, Twitter said the account did not exist. I was satisfied. A few hours later, after spending some time thinking about Twitter’s spam problem, I noticed that it was back, perhaps unsuspended (or possibly there was some error when I reloaded it).

I can appreciate that the anti-spam people at Twitter have a difficult job (this report says they once accidentally blocked Biz Stone’s and Evan Williams’s accounts). They are trying to craft algorithms to deal with a constantly adapting and increasingly sophisticated enemy. Furthermore, Twitter has only around 500 employees; it can’t possibly have more than a couple of them working on anti-spam measures.

Machines are notoriously bad at pattern recognition. If only Twitter had millions of human-level intelligences who could ease the burden on the computer algorithms. But of course, they do! If Twitter’s users could somehow be persuaded to use the “report spam” function consistently, Twitter’s spam problem would be much reduced. And for all the ingenuity that the spammers have shown so far, it’s unlikely that they would ever be able to beat the human filter.

Reporting spam has, in addition to the private revenge component, a public goods component. There are positive externalities to being a good online citizen, and therefore good online citizenship is underprovided.

Fortunately, since Twitter owns Twitter, it can solve this externalities problem. Coase shows in his article The Lighthouse in Economics that lighthouses (the canonical public good) were provided by private actors because they were affiliated with harbors. Harbor owners undertook the building of lighthouses because well-lit harbors were more profitable. Public goods problems can be solved through the good incentives that ownership provides.

The solution to Twitter’s spam-reporting public goods problem, therefore, is entrepreneurial innovation on Twitter’s part. Twitter needs to incentivize its users to make use of the report spam button every time they receive a spammy @reply or are followed by a spammy account.

My incentivization suggestion is to increase, for users who report accounts that are ultimately verified or not contested to be spammy, the likelihood of showing up as a recommended user to follow. Publicize this change in the account suggestion algorithm and overnight, millions of Twitter users will report spam every chance they get. What is more, it will turn the spammers against each other. The spammiest users will make extensive use of Twitter Search to locate other spammers to report them, to increase the likelihood of being a suggested account.

If this suggestion were implemented, it’s hard to imagine how spam could continue to be a problem on Twitter. The ingenuity of the spammers will be matched by the ingenuity of people who are trying to game the Twitter account suggestion process. And if Twitter is going to recommend accounts, why not at the margin recommend those who are good online citizens?

If anyone at Twitter is reading this, I am happy for you to use this idea without further permission from me as it will solve an aggravating problem for me and other users. But only the first one is free. I have an economist’s perspective and a number of other ideas; if you want them, I’m going to need some stock options.