It's a trade-off. You lose some clarity (how do I know what params to pass) and checking (did I pass the right number of arguments?) and gain flexibility (the method can default params it doesn't receive, we can deploy a new version that takes more params and break no existing code)
You can see this question as part of the larger Strong/Weak type discussion. See Steve yegge's blog here. I've used this style in C and C++ in cases where I wanted to support quite flexible argument passing. Arguably a standard HTTP GET, with some query parameters is exactly this style.
If you go for the hash appraoch I'd say that you need to make sure your testing is really good, problems with incorrectly spelled parameter names will only show up at run time.
