Hacker News Reader

Comment by wildekek | Hacker News Reader

How do you weed out false positives?

Bingo. It seems pretty easy for one restaurant to file reports about a rival.

I'd be happier with a site that aggregated food poisoning reports from various government health departments, but I'd worry about Joe Random telling me he died eating someplace.

But then it wouldn't be crowdsourcing, which is the point.

We do a pretty deep dive on every report in the moderating process. Someone with malicious intent could get one report past us. Getting multiple reports past us would be do-able but harder. Getting them actioned by public health would be close to impossible. Rolling reports off after 30 days hopefully de-motivates trolls, if their hard work is gone so fast.

This wasn't much of an answer to "how", just restating that "we have a way to do it, we promise!". Could you give some explanation as to why it would be close to impossible to create enough reports to have them actioned? What markers do false reports have? It seems you only collect some free form text, and I can't imagine any way you could tell a good fake from a real report. So to a business, driving out a competitor out may very well be worth the cost of a few unscrupulous internet warriors.

From us the data to goes to the governing public health agency. Each spammer would need to individually re-present their case to that agency including a 72 hour food history, and potentially bio testing. The next steps can include epidemiolist, site visit, lab testing and more. The likelihood that everything could be falsified including lab tests with all these government officials is low, and it is only when a case passes through this process that a health agency will consider making an announcement such as "Restaurant X is the source of X pathogen causing X people to be sick"

We collect more than free form text.

Thanks

So if I went out to eat and got sick, I might think I'm performing a public-service by reporting it.

Then you come back to me and say "Hey your report was weird, go get some lab-tests done"? That seems like a situation which would deter most reporters.

Once we receive and moderate your report. It will go to public health. There are many paths it can take from there. We don't require lab tests, that is just a potential scenario, most common where there is either an important (deadly) pathogen potentially involved, or if it appears to be potentially a larger outbreak, and it is a pathogen that can't be detected in the facility. It's not necessarily ever a requirement. What we offer to consumers is that if you speak up there is potential that your report can make a difference.

Thank you for the clarification.

Also very few goverment agencies release complaint data, and for those few that do, it is not done in a timely actionable accessible way.

Depends on what you mean by weed out. Each report is a signal. A single report on our site is a weak signal but still may result in a) an inspection and identification of critical violations b) discovery of an outbreak c) a false positive. One hundred reports relating to the same place same timeframe, would be considered a strong signal, and more typically results in a) and b). Our approach to handling the false positive aspect is that we roll data off our website after 30 days.

Reply on news.ycombinator.com