Data does not really speak for itself and experts are no less susceptible to bias than anyone else.
It is true that statistics can be manipulated, and experts are subject to bias like everyone else. Yet, science itself is proof that a preponderance of evidence, and the discussion of experts, are sufficient to continuously improve understanding and create reliable structures. If experts and data were truly unreliable, engineers would be as unable to build sturdy bridges as they were two centuries ago, and physicists wouldn’t have known how to develop the laser in your hard drive. The experimental method isn’t flawless, but it’s better than passing a law that has not been tried experimentally, first.
surveys don’t reveal peoples real preferences
It’s true that there is a disparity between a survey and observed behavior, in many ways. Democracy itself is just a survey, so that disparity applies to the vote just as it would to the surveys I propose. Yet, in a technocracy, peoples’ behavior would be data available to experimentation, as well. Technocracy has surveys, and use-cases. Democracy is only survey, without ever measuring real behavior.