> Smoking causing cancer was substantiated by some of the strongest correlations in medical science history.
Of course. As westoncb pointed out, the actual risk of smoking isn't what I'm talking about. The visible damage from smoking is only seen later (i.e. when symptoms of cancer or emphysema start to appear), so it is extremely difficult for the average person to properly evaluate the risk. Humans use approximate pattern patching and heuristics when evaluating risk; this fuzzy approach to input data saves energy but fails spectacularly in some situations.
> I'm not sure where you'd observe something anything strong like that with regards to surveillance
You're exhibiting the exact problem I'm talking about: you are only considering current problems. There may not be any problems at the moment for you to observe. While that's great for today, the databases that contain an increasingly accurate profile about your activities still exists in the future. You are (severely) underestimating the risk of your data accumulating in (approximately) permanent databases if you haven't included things like:
* A judge during sentencing or parole hearing using a model that claims you are high risk because social media posts you made today correlate with the "high risk" group de jour.
* Your religious, ethic, or political group becomes the next scapegoat. Previously recorded data allows the members of that group to be easily identified "for public safety".
* On the business side, your future insurance company raising your rates because prohibited information (e.g. ethnicity, income) could be inferred from the data you are currently generating.
Those are just a few of the obvious risks that we know about. As time progresses, more risks will be discovered.
> "addicted"
That aspect of smoking is entirely orthogonal to my point. Smoking is merely one obvious example of poor risk estimation.
> It's an absolutely horrible thing you start doing to your body immediately.
Yes. It's a damaging process that wont be easily visible to the average person when they are subconsciously judging the effects of their previous cigarette and if they should buy more. Over the last few decades we have had some success at educating people about the actual risks to counteract these common errors in perceived risk, which is exactly what we need to do for data sharing and the surveillance problem.
Of course. As westoncb pointed out, the actual risk of smoking isn't what I'm talking about. The visible damage from smoking is only seen later (i.e. when symptoms of cancer or emphysema start to appear), so it is extremely difficult for the average person to properly evaluate the risk. Humans use approximate pattern patching and heuristics when evaluating risk; this fuzzy approach to input data saves energy but fails spectacularly in some situations.
> I'm not sure where you'd observe something anything strong like that with regards to surveillance
You're exhibiting the exact problem I'm talking about: you are only considering current problems. There may not be any problems at the moment for you to observe. While that's great for today, the databases that contain an increasingly accurate profile about your activities still exists in the future. You are (severely) underestimating the risk of your data accumulating in (approximately) permanent databases if you haven't included things like:
* A judge during sentencing or parole hearing using a model that claims you are high risk because social media posts you made today correlate with the "high risk" group de jour.
* Your religious, ethic, or political group becomes the next scapegoat. Previously recorded data allows the members of that group to be easily identified "for public safety".
* On the business side, your future insurance company raising your rates because prohibited information (e.g. ethnicity, income) could be inferred from the data you are currently generating.
Those are just a few of the obvious risks that we know about. As time progresses, more risks will be discovered.
> "addicted"
That aspect of smoking is entirely orthogonal to my point. Smoking is merely one obvious example of poor risk estimation.
> It's an absolutely horrible thing you start doing to your body immediately.
Yes. It's a damaging process that wont be easily visible to the average person when they are subconsciously judging the effects of their previous cigarette and if they should buy more. Over the last few decades we have had some success at educating people about the actual risks to counteract these common errors in perceived risk, which is exactly what we need to do for data sharing and the surveillance problem.