If you want to detect how tall your target is and tailor the drop height to suit, then you are going to need more precise sensors and control than the other (excellent) answers here.
One option might be to use a Sonar Range Finder as suggested in this answer to What is the cheapest / easiest way of detecting a person? (Thanks Ian)
Basic idea
While the SRF is detecting a distance of more than your maximum drop distance, the 'spider' would lie in wait. If the SRF detects anything closer, you immediately start the motor unspooling the cord and the spider will start dropping.
As the spider is dropping you continue monitoring the SRF and stop unspooling when the length of cord you've unspooled is approximately the same as the minimum distance measured by the SRF since it started dropping. This should place the spider right at eye level of the tallest person there, for maximum effect.
Finally, after a short delay, start the motor slowly returning to it's rest position by spooling the cord back in and resetting for the next attack.
Complications
You will need to be careful about the weight of the spider. It needs to be heavy enough that drag (air resistance) does not significantly impede it's acceleration due to gravity, but not so heavy that it could harm anyone or overload your motor.
Even if your spider isn't too heavy, you will want to make sure that there is 1.5 to 2" of foam between the heavy centre and anything that might touch anyone. You will also need to make sure that there are no stiffening wires or springs which can end up protruding out of the spider as it wears. All basic health and safety stuff really.
If the spider is dropping into the SRF sensor beam area, then obviously you will not be able to improve your initial estimate of drop distance, and will just have to drop to that first range measured. This might mean that you end up dropping to teh height of someone's hand rather than to the level of their head though.
If you want to be clever, you can arrange your drop trajectory to be controlled so that you never have too much loose cord, this will minimise the risk of the cord getting tangled and snagging on anything. In this case you will want to ramp up the speed of the motor such that the cord is being spooled out at the same rate as the acceleration of the spider due to gravity, i.e. $9.8 ms^{-2}$.
You may or may not also want to control the deceleration rate as you approach the desired drop length. The faster you stop it, the more dramatically the spider will bounce and jump around, but the more stress you will put your system under.
Finally, you will want to size your motor such that it can move relatively quickly under light load (when dropping) but still have the torque to stop the spider at the end of travel (and winch the spider back up).
Some calculations
The formulae for uniform acceleration is $s=ut+\frac{1}{2}at^2$ where $s$ is displacement, $u$ is initial velocity (zero in our case), $a$ is uniform acceleration and $t$ is time.
Assuming that you have a spool of 10cm in circumference, and want to drop the spider at most 2m then $s=2$, $u=0$, $a=9.8$ and $t$ is unknown: $t=\sqrt{\frac{2}{\frac{9.8}{2}}}=0.638$. To spool 2m in that time seconds would require the spool to turn at almost 2000 rpm on average ($200/10/0.638*60=1881$).
This seems a little unreasonable, so doubling the spool to 20cm circumference would drop the motor speed to under 1000rpm on average, but we are now looking at quite a bulky bit of mechanics.
Conclusion
If you get the design right, you might even be able to fit all of the components needed (spider, spool, motor, drive electronics, control electronics, sensor and power supply) into something which you can screw into a light fitting and thus hide in a (large) lamp shade.
