You've identified the issue in the example: risk aversion, which is often not modeled in 'classical' econ, e.g. not in auction theory generally. In Scenario 1,

You've identified the issue in the example: risk aversion, which is often not modeled in 'classical' econ, e.g. not in auction theory generally. In Scenario 1, you have a choice between definitely not dying (buy the antidote) and dying with 1/1000 chance (don't buy it). They're asking how much you'd pay to switch from the second to the first. In Scenario 2, you have a choice between definitely not dying (stay home) and dying with 1/1000 chance (go to the lecture). They're asking how much you want to be paid to switch from the first to the second. If your utility is quasilinear in money (meaning it equals utility for dying plus final quantity of money), as in 'classical' econ, then these decisions are the same assuming you don't care about attending the lecture.