There comes a point in a major lottery like Powerball where the expected value of buying a ticket technically becomes positive. However, the odds of actually hitting it remain so remote that the likelihood of coming up empty before you die or stop buying tickets (whichever comes first) remains too great even though the EV is positive. I forget what the statistical term for this is, but the math ceases to be practical.
Here, it's +EV to roll the dice compared to taking the sure $100K. However, you only get one shot, and 5 out of 6 times you're going to get nothing.
For someone who thinks $100K isn't much money, they still think rolling the dice is the smart bet. For someone poorer, for whom that $100K would be anywhere from life changing to seriously useful, it may be more worth their while to just take the $100K. The alternative leaves them still broke far too often to justify taking the shot.
Edit: Though it's not the term I couldn't remember above, the idea does tie into another statistical concept whose name I do know: Risk of ruin. It's from gambling and is essentially the probability that a skilled gambler with a suitable bankroll might lose often enough to lose their entire bankroll despite their positive expectation.
There is a relative risk of ruin here. The rich man for whom $100K isn't much money doesn't really have any risk of ruin to consider. He's got assets either way. Go ahead and roll the dice.
However, the poor man for whom $100K is a big deal is facing an 83.3% risk of ruin, i.e. the probability that they don't roll a 6 and continue to not possess at least six figures of cash. For them, the risk of ruin chasing the billion is simply far too great to justify not taking the sure $100K