Google Maps must have pur­chased a new suite of road infor­ma­tion recently. Or maybe they algo­rith­mi­cally tried to “improve” the dataset they had. Used to be it knew local geog­ra­phy pretty well; now, not so much.

When ask­ing for direc­tions from our farm (on Walsh Road, Web­ster Town­ship) to the Dairy Queen in Ham­burg Town­ship, the dri­ving (not walk­ing!) algo­rithm sug­gests we stay on north­bound Scully. If you saw it from a dis­tance, the satel­lite image would lead you to believe that, yes.

Except that many years’ fierce argu­ment at the county bor­der has left a nasty but potent gate block­ing the road, which will per­sist into the fore­see­able future.

If you were to drive up the rough, mainly untended Scully Road on a snowy day, try­ing to get (say) to a hos­pi­tal in Pinck­ney or some­thing, the least that would hap­pen is you’d waste a half-​​hour try­ing to back out of the last few hun­dred yards with­out end­ing up in a ditch… once you arrived at the impass­able gate at the bor­der, and well after you had tres­passed on a pri­vate road at the end.

The De Lorme Michi­gan Atlas & Gazetteer, a nice old printed book I keep in my car, and which is so obso­lete that it shows lit­tle red lines for roads of all sizes and char­ac­ters, man­ages to catch the gap.

Now every dataset con­tains errors or miss­ing infor­ma­tion. But every time that dataset is used to make a sin­gle, sum­mary state­ment, based on a sin­gle model? Bad­ness can hap­pen in unex­pected ways. In fact, I am obliged to be cur­mud­geonly about it because of my pro­fes­sional expe­ri­ence in these mat­ters: it is always wrong to present a sin­gle answer for any multi-​​objective or highly con­strained decision-​​making prob­lem. Big, fat period.

I can’t com­plain, in all hon­esty, about advice given by a black box operations-​​research algo­rithm that on inspec­tion I knew was incor­rect. You get what you pay for. But I can com­plain about a cul­ti­vat­ing a mis­lead­ing user expe­ri­ence in a ubiq­ui­tous data-​​driven deci­sion sup­port sys­tem that presents only one solu­tion at a time to the decision-​​maker. Hell, every iPhone in the world has one of these on it; they’re all wrong, too.

No, I don’t think I am feel­ing lucky, Google. And you didn’t even ask.

I want to see a sheaf of routes. The lit­tle “adjust the route and recal­cu­late a new one using my mile­stones” han­dles Google intro­duced a few years back are a beau­ti­ful thing, a cun­ning arti­fact and a use­ful tool! And of course, the stan­dard “avoid­ing high­way” or “fastest” tog­gles let me reach in and fid­dle with the search method. But only indirectly.

I want the objec­tives right there, not com­bined. I want not just to sur­face the meter (to use a phrase Dan Cooney’s taught me), but sur­face all of them. I want choices cou­pled to clearly dif­fer­en­tiable sup­port­ing arguments.

Like the basic Google Search results them­selves: ten routes at a time, ranked some­how. Or not even ranked, but handed to me as a Pareto-​​equivalent set of alter­na­tives, some faster, some bumpier, some with big­ger roads, some with more gas sta­tions, some more scenic. Heck, maybe I just want to know there are at least ten ways to go back and forth, so I can stage a race, or not get bored on my com­mute, or defend against unwanted SUV inva­sion by a for­eign county or something.

At least some­times. Stop assum­ing I’m feel­ing lucky.

Next time, we can bitch about the mis­lead­ing user expe­ri­ence and illu­sory author­ity cre­ated by the fuckin’ weather fore­cast for­mat. Every­body com­plains about the weather fore­cast, but nobody does any­thing about it.

Lather, rinse, repeat.