> So the problem is that you can make some statistical predictions about what someone's views are based on their reading habits. But you can't morally judge them on that basis
Companies are making these kinds of flawed assumptions about you and every one of us every single day. They often then sell that info to data brokers who happily sell that data to others who will then start with incorrect assumptions about you as an individual and let that influence how they interpret the rest of the data they collect about you.
It's a real problem because those flawed data sets you aren't allowed to see, contest, or update are increasingly being used to meaningfully impact your everyday life in ways that you'll never be aware of.
The bottom line is that companies don't care. If they will make more money by being right most of the time that's what they are going to do. You might be the fittest, healthiest person on Earth, but if your health insurance company sees that people in your area code have started buying fast food more often they can decide to raise your rates. They won't tell you why they did it. They'll just do it. You could be the most financially responsible person on Earth but if you live in the wrong zip code don't be surprised when you get denied certain services or told that a company's polices are one thing when they would have told you they were something else if you lived on the other side of town.
That said, it's probably a whole lot easier to make accurate predictions about people than you think. Sure you'd read fox news, but most of your time is probably not spent on right wing sites, you probably aren't leaving comments espousing right wing talking points, and you probably aren't donating money to right wing causes.
With enough data it's not that hard to figure out if you're regularly hanging out at stormfront because you're working for the Anti-Defamation League or because you're a racist.
Algorithms can detect (and exploit) mental illnesses like bipolar disorder and Alzheimer's companies can certainly detect "stupid" well enough for their own needs.