You don't bet if it's going up or down.  No one is going to bet that it's going down, of course.  The idea is no one really knows what the difficulty will be in December.  Let's use a block 10 increases in the future from now as an example.  If I'm a miner, my profitability will vary tremendously based on what that difficulty is (along with every other along the way).  I cannot make an accurate assessment of whether my investment is profitable.  However, you can have difficulty futures, such that if the difficulty is above X it pays one party, and if it's below it pays the other party.  Both parties put in money up front, and get paid out in December whatever the results are.  It does not matter what X is, there is a price that someone could theoretically pay for both sides.

So you have various levels of X, and they require different contribution shares from each party.  The winner of each contract wins 1 BTC.  For example, you may have:

Difficulty	Price to win if over	Price to win if under
500,000,000	.01BTC	.99BTC
400,000,000	.25BTC	.75BTC
300,000,000	.60BTC	.40BTC
200,000,000	.90BTC	.10BTC
100,000,000	.99BTC	.01BTC

The prices will roughly track the probability of the difficulty being at that level.  For example, with these made up numbers, it would infer there is a 1% chance of being > 500M, a 25% chance of being above 400M, a 60% chance of being above 300M, a 90% chance of being above 200M, and 99% chance of being above 100M.  These kinds of prices will allow him calculate how much BTC he should mine in those 2016 blocks based on current expectations.  In addition, he can also buy some of these contracts, such as above 400M and above 500M, such that it will compensate him if difficulty rises too much, he can at least recover some of his investment.  This has a cost, but if it is lower, he will make up for it in extra bitcoins mined.

Why would someone take the opposite side?  Profit.  Say I think the difficulty will only be above 400M 10% of the time.  So I can bet 75 BTC against it going above 400M.  10% of the time, I lose my 75BTC.  90% of the time, I win 25 BTC.  This means I profit 60BTC.  I still win even if the difficulty rises to 399M.

Taking it an additional step forward, someone could offer a more complicated contract such that the miner only needs to pay a simple price, and the result of the payout would be proportional to the difficulty.  Say a miner only wanted to hedge above 400M in this case.  He could be paid Expectations of Mining at 400M - what he would make mining at the actual difficulty.  So if it was 405M, he would get a small payout, and if it was 600M, he would get a larger payout.  He would pay a small amount for this service, and would use it to make his expectations much more known in the present and make his decision making on expanding rigs, etc... easier to plan.