Any time one moves against a force, work is done, as per work = integral force over distance times the cosine of the angle between the vectors. What this means in context is that when the man changes position (Displacement), work is done against centripetal acceleration (Force). If the man moves radially outwards, the angle between is zero, and thus the work done with centripetal acceleration adds to the man's kinetic energy.
In practice, to return to his initial position, work must be done; that work doesn't disappear, it simply changes forms. Some of the work done as the man moves out may be lost as friction(he has to slow down to prevent from flying off). When he pulls himself back to the center, he must use energy, most likely in the form of stored chemical energy (his muscles).
If anyone feels like doing the math, the merry-go-round would be approximated as a spinning disk with
KE = 1/2 I [itex]\omega[/itex]2
P = I [itex]\omega[/itex]
(I is moment of inertia, [itex]\omega[/itex] is angular velocity)
Where I for a disk is 1/2 m r2
The man is a point mass on the disk with
KE = 1/2 m V2
P = I [itex]\omega[/itex]
I = m r2
V = r [itex]\omega[/itex]
Realize then that angular momentum is conserved, IE: Initial Total Momentum is equal to Final Total Momentum. The change in kinetic energy of the system then turns into a different type of energy.
Alternately, to simply find the work done by the man moving, you could integrate centripetal force through the (radial) distance traversed by the man. Of course, to do that accurately you'd still have to calculate a changing value for [itex]\omega[/itex]...
Unfortunately either approach becomes rather complex when you realize that both [itex]\omega[/itex] and r change with distance. I think it's solvable, but I'm not feeling like it atm.