,
11 tweets,
2 min read
Read on Twitter
1) The other night was our wedding anniversary. There’s this Japanese place to which I wanted to take Mary for dinner. I called to reserve; no answer, but an answering machine took the call, and the outgoing message included a promise to get back to me later that day.
2) We didn’t get a call back, but since the reservation was for 6:00pm, we didn’t anticipate a problem. According to Yelp, that's when the place opened, and 6:00pm on Tuesday is usually not high-demand time for Toronto restaurants. So we headed down. The restaurant was closed.
3) There was nothing to indicate why. It was Tuesday evening. Yelp suggested that hours were 6:00pm-11:30pm, and that it should be open. There were no opening hours posted at the entrance. Nothing on the restaurant's Web site indicated any reason for the restaurant to be closed.
4) On Wednesday evening at 7:00pm, I got a call from someone at the restaurant asking if we would be honouring our reservation for 6:00 (which they had never called to confirm). I explained that I had called Tuesday, had showed up, and had been disappointed.
5) The person on the phone told me that the restaurant was closed Mondays and Tuesdays. I identified all the places where this information was missing (all under the restaurant's control) or misleading (on Yelp, crowdsourced but evidently not verified by anyone).
6) Now: you might be thinking that this is a complaint about customer service. It is, a bit, but not much. Instead, it's a warning about assumptions, heuristics, and about the illusion of reliable data. My assumptions and heuristics failed here; but data on the Web didn’t help.
7) Machine learning "is becoming powerful because we have access to all this data". Which would be great if that data were valid and reliable, but much of it is crap, *especially* when data gathering is crowdsourced without being evaluated critically. And data gets stale, too.
8) I asked the person on the phone to notify management about posting hours on the Web site and on the front door. There's no telling when or if those postings will happen. I wasn’t motivated neither Tuesday nor Wednesday, to notify Yelp about the inaccuracy. I might do that.
9) My point here is that the Web itself is massively crowdsourced, prone to being inaccurate, imprecise, incomplete, inconsistent, contradictory, and out of date, to some degree. Not all wrong, but difficult to determine what to trust and to distrust, and when.
10) Human physical capabilities can be extended and enhanced and enabled and accelerated by machinery, although sometimes with problems and risks. Problems with machines that do physical things are relatively easy to observe. With machines that mediate social tasks, not so much.
11) Machinery neither develops nor applies tacit knowledge, wisdom, or responsibility. We must do that, lest we we disappoint, harm, or even kill people. We must be careful that our ambition doesn’t dazzle and distract us. Today's tiny Yelp problem is tomorrow's fatal car crash.
