I only went to Part 1 of this talk; I had been planning to go to a different one but it filled up before I got there. This was my second choice but it was a very nice reminder of the most important points about training behavior chains. I wish I had stayed for the second part, actually!
chain with external cues - external stimuli provides cues (ie handler giving signals)
chain with internal cues - behavior or environment provides cues
ingredients of a chain: flient behaviors (solid cues with good stimulus control), positive trained cues, well-timed cue delivery (just as with well-timed clicks)
your cue is your click!: timing matters!; poorly timed -> will not reinforce previous behavior; behavior that is not reinforced is not maintained - chain will break down
demo: volunteer got up and was cued to spin, then clap, then put her hands up
first with external cues, then with internal cues (she was to decide on her own when to switch to the second and then third behavior)
the behaviors started to meld, degrade, become sloppy - this happens with all animals
beware of poisoned cues!
does not reliably predict positive outcome (could be R+, P+, etc)
common problem in many behavior chains
a poisoned cue is not a tertiary reinforcer - it will not support the previous behavior
without internal reinforcement, the chain will break down
backchaining
best choice for predictable chains
always working toward more fluent, more reinforced behavior
minimal confusion, maximum enthusiasm
cheer completed task, not an attempt - avoid aborted attempts, mistakes, R+ of incomplet chain
never fix a behavior within the chain - split it out to fix individually
No comments:
Post a Comment