Bas Vodde posted an article on action planning in retrospectives . It's a tough issue, and I agree with Bas' take on it. Team members need to see clearly how the actions they choose will affect their work long term. Bas suggests each proposal for action links each near term action with the long term goal it will help the team achieve.
Example
Long term goal: Balance development and maintenance duties among team members to avoid burnout.
Action Next Iteration: James will track total team time and individual team member time spent on development and maintenance tasks. Bring tracking chart to next retrospective.
(Bas gives another example in his article.)
Of course some teams need help in selecting only a few, do-able actions. Bas describes an action prioritizing activity that has worked well for teams. In my experience similar activities that move from individual idea generation to pair selection to quad selection, and so forth, work quite well with teams of ideal Agile size, (7 +/- 2), and a bit larger.
With larger groups, more than 18-20, it becomes too time-consuming and tends to favor the "least common denominator" action, rather than the inventive, audacious ones. The larger the group, the more the dominant individuals on the team can influence the outcome. Bas mentions the challenges of trying the activity with a group of forty.
Thanks for sharing your ideas, Bas!