Some kinds of trading, including high-frequency trading, add liquidity; participants that do this are called market makers, and market makers reduce volatility by making it cost more to move the price.
Apologies, but you are mistaken. Market makers, in order to qualify as such (and receive the official designation from an exchange) have to satisfy certain condition that are not satisfied by HFT operations. The relevant ones for the 'provide liquidity' part are: to always maintain bid/ask quotes and stand by to execute trades at the quoted prices. There are rare exceptions when one can pull off, but that's the gist of it. Bit of a quaint notion nowadays, what with multiple connected electronic venues, but some exchanges still have it and it does use electronic quoting nowadays. However, this little detail does disqualify the majority if not all HFT shops from market making in the conventional sense.
Now, the good part that fast electronic trading does (which is not exactly the same as HFT) is pricing arbitrage between exchanges. Securities listed on multiple exchanges and/or in dark pools often enough get crossing prices, where one can arbitrate the difference and re-bring quotes in line. For example, when a price starts moving on one exchange (due to a block trade, let's say) price propagation is something that nowadays happens a lot faster. Similar things can be done with arbitrages between underlying instrument prices and derivative ones. If you choose to include this in HFT, then I'll grant you it's a good thing and it does improve liquidity[*]. It still does not make HFT players market makers, but arbitrageurs are good for price discovery[**]. However, you can't quite separate this part from the *ahem* 'evil' one, as you called it. It's simple market opportunity taking. Well, at least you cannot unless you start requiring a set of conditions that would make it unprofitable to engage in the behaviour you want to prevent.
Finally, going back to my original post, please do follow the link in there. It's not about 'the' flash crash, but one of the many single-stock flash crashes that happened since. This one was on GOOG, which is not an illiquid stock, and lasted less than 2 seconds. That's quite different from fat-finger crashes where extra-large orders are mistakenly placed and clear all the bids in the books on their way to the great below. (and btw, I do wish that my posts on this topic would stop getting irrelevant replies about the SPY flash crash when I'm not referring to that one and not even the same type of phenomenon)
[*] by decreasing quote spreads, which is the main proxy used for liquidity measuring. Still, I am yet to see convincing evidence for the part HFT played in this as opposed to the general improvement in connectivity between markets.
[**] However, having an option quote react faster to the price change of the underlying does not necessarily mean I'll get more liquidity for it. The liquidity question is thus tricker than it seems.