I've been asked more than once why we should bother sending humans into space to do the work at which our robotic landers and rovers have been so successful. The answer is: because they actually aren't terribly successful.
The probes we send to other planets have a very narrow set of tasks they are able to perform. The Viking landers could take pictures of stuff. Spirit and Opportunity can roll around, and they can drill things. The Phoenix lander can dig a bit, but it can't move. Our latter Mars missions also have a bunch of fancy equipment for measuring flux capacitance and theta radiation. Which is good. We can learn a lot from that.
All of those pieces of equipment which are specially designed for use by the robots, however, can easily be stocked into a small human laboratory. Plus a lot more bulky equipment. Large centrifuges and whatnot. I hear scientists enjoy those. Need a soil sample? Send Johnson outside with a trowel and a Ziploc baggie. Five minute process. See something interesting a few miles away? Hop on a little rover and go take a look. A few hours at worst. Basically, if there is interesting science to do, just do it.
Robots can't particularly match that versatility at the rate we're currently progressing. Robots must be told what to do at every step of the way, so that they don't become permanently stuck in a sand dune or run out of juice during dust storm season. This is a little annoying, because signals between here and Mars take between three and 20 minutes one-way. So, if you're navigating a robot through a dense rocky field, and you tell it to move about ten feet per communication and then wait for confirmation that you haven't driven the robot off a cliff, your average speed is at best 1.7 ft/min and at worst 0.25 ft/min. That's lame. I'd just jump over that shit, because I'm a human, and I'll do as I please.
Of course, we can eventually wring a bunch of interesting science out of the rovers and landers. However, we can only learn as much as we think we already know. At the cost of tens to hundreds of millions per robot, you only invest in robot science that you know will find something definitive, like an answer to the question of water on Mars with the possibility of ancient life. The rover isn't keeping its eyes pealed for interesting and strange phenomena outside the realm of its narrowly-focused mission. Humans, by nature, excel at that. That's why we have things like antibiotics and radios. Science born of random observation ought to be sustained, even at high cost.