Interesting discussion on adverse selection and uninformed order flow. I have done a little work in this area and the markets have changed a lot. Here is the short version of what I have seen: There is a way to measure this by decomposing the spread on a trade into the price impact and realized spread, it is probably in the Harris book or not too hard to find on the web. The price impact is how much the quote midpoint is moved by a trade, and is a measure of information in the trade or adverse selection. The realized spread is how much of the spread is captured by a "dumb" market maker after taking out the price impact. In theory, and I understand in real life in the past, the spread is set wide enough to cover the price impact and then some. Not on every trade of course, but on average over reasonably large samples the mm should more than make back on the spread what is lost on trades where he is picked off. But in a large amount of recent data on individual stocks and a couple of the more liquid futures contracts I have looked at, this is no longer true. Realized spreads are negative. So roughly speaking an mm would lose money buying at the bid and selling at the ask on every trade in the markets I have looked at. I have some guesses why this is the case but I am curious if anyone knows or thinks they know why. This is ignoring rebates, but not all these markets pay rebates. This must mean the mm's are now either very good at knowing fair value or differentiating uninformed orders to trade with or both.
Actually there are and competition to trade with them can be fierce. Payment for order flow is based on their existence, and so are a lot of the dark pool rules. But if you meant less of them make it onto an exchange to find a limit order placed by a retail trader I would agree.
In response to the OP, I'd work the order if it was large, otherwise I'd just give them a wide quote. This will prevent losses.