You're actually very close.
The null hypothesis (Ho) is want you want to try to "reject" (you want to try to find enough evidence to say that Ho probably isn't true, so you reject it in favor of Ha).
Think of it in terms of the court system - "innocent until proven guilty beyond a reasonable doubt." Ho represents "innocence." Rejecting Ho in favor of Ha represents "reasonable doubt" that Ho is true.
In your problem,
Ho: mean response (placebo) = mean response (control)
Ha: mean response (placebo) < mean response (control)
I use the "<" sign because of the problem wording:
"They hoped to find that the placebo reduced the mean response"