HFT is a great example mentioned by a sibling comment. You kind of need some number of traders out there to make sure that market prices are honest and fair. HFT has done a great job of reducing (a) the number of people involved in trading per dollar traded and (b) the amount of money made by traders per dollar traded.
To be clear, nothing wrong with that. But from a personal perspective you start to question your contribution.
But how does eg an algorithm that sells stock when the newspaper headlines have a "negative sentiment" help ensure that "market prices are honest and fair"? What's "honest and fair" about stocks across the board dropping because of bushfires or a virus outbreak, only to rise back to pre-headline levels half a day later?
When the virus starts, the markets adjust downwards because there is a chance that virus kills tens of millions, which would have severe economic impact.
As the world public health starts to reduce the error bars or bad outcome likelihood, the market adjusts.
The problem is that you’re thinking of something as having a concrete value. Instead, think like Tetlock teaches in superforecasters. Then you will have a better model of what’s going on.