Regression to the Mean - Negative or positive reinforcement for skills

[Doug]

(moved to the Practice area -- ToddG)

Thought this video was very insightful. Starts with which is better, positive or negative reinforcement in performance, then goes on to discuss the statical ramifications of the interpretations. Definitely has application to shooting skills; at least for me.

http://www.youtube.com/watch?v=1tSqSMOyNFE&feature=youtube_gdata_player


http://www.youtube.com/watch?v=1tSqSMOyNFE

[1911Nut]

Thanks for sharing this. Very interesting.

[GJM]

Very interesting. As a psychology major, BF Skinner fan boy, and veteran of pigeon lab, I learned some stuff.

So, is the quest for consistency, when trying to do something anywhere near your personal best/record level, a fool's game?

[YVK]

There is a bunch of wrong logic in this video. Results of an activity performed by trained people will not follow the same distribution as classic Gaussian curve of results of random folks, because trained folks is not a random group anymore. Mean values for these two groups will be different. One's performance may randomly jump around the mean, but that's why we look for trends and p values, so we can apply feedback affecting non- random patterns.
I have very minor insight in psychology, but my personal observation is that positive reinforcement helps to train good performers and promotes trainee's retention, but not as helpful in training great performers and testing a commitment as negative feedback is. I may be speaking out of my ass here though.

[Joseph B.]

I think positive vs negative enforcement/correction is individual based. Some people do respond well with negative response to their performance. I would say a lot more do not. However, the negative reaction to their performance, normally will burn the event into their memory, thus the right way to do something is remembered due to the negative response to the bad/poor/wrong performance.

My soldiers did not always respond well to negative correction, but the majority did to some point. Once that point has been hit, I normally would have to shift gears to positive enforcement to get continued performance, however, at some point the positive enforcement would fail to drive performance, and I would have to shift back to negative corrections. The cycle would continue, and was very much dependent of the individual. I think mainly, understanding how the individual is perceiving the message and reacting to their perception, neg vs pos, and which is gaining the performance direction you as the trainer or leader desire.

Not saying “this is how it is” just my personal observations in trying to get performance from a mess of different people, from different backgrounds and cultural upbringing.

It is an interesting video nonetheless.

[GJM]

As we learned feeding the pigeons in pigeon lab, continuous positive reinforcement will become progressively less effective. The key is to use an intermittent reinforcement schedule -- makes the pigeons peck a lot harder since they never know when they will get a kernel.

[ToddG]

So, is the quest for consistency, when trying to do something anywhere near your personal best/record level, a fool's game?

I think it depends on what you're practicing and how. Looking at it from the perspective of the video in the OP, the more random your results the higher likelihood that you'll fall well below (or above) your mean on a given rep. It's a basic standard deviation (σ) issue.

Let's take two shooters, Tom and Gerry. Both have an average El Prez score of 10 As and 2 Cs in 8 seconds.


Tom is very consistent. He always shoots between 9 and 11 As and he always shoots between 7.75 and 8.25 seconds. He's the shooting world's 6σ master.
Gerry is far less consistent and shoots anywhere from 6 to 12 As and from 6 seconds to 12 seconds.


If everything comes together perfectly for Gerry on a particular run, he could get a 6s 12A El Prez run. Tom's not going to be able to beat that. But Tom's never going to do worse than 8.25s 9 A, while Gerry turns in, on a worst case run, 12s and just 6 A.

They both have the same average performance over time, but Tom's is more consistent. He won't have flashes of brilliance but neither will he crash & burn on occasion.

The question then becomes, how do you move the average toward a better result?

Think of it like resetting a trigger on a gun that has a 0.155 reset. There's a certain degree of imperfection in human movement that prevents you from moving your trigger finger exactly 0.155000000" each time you fire a shot. Going 0.154" will result in abject failure. There are two approaches you can take:


accept your current σ and simply move the target (reset point) forward so that you'll almost always be within an acceptable envelope to reset the trigger, or
reduce your σ to near-zero so that you'll almost always be within an acceptable envelope to reset the trigger.


From a human performance standpoint, a combination of the two works best. Riding the reset right to the edge still always puts you at risk for short stroking but purposely releasing the trigger all the way forward means more finger movement than necessary. So you want to improve both. You want to move your goal (reset point) to something you can achieve with a very high level of success on demand but you also want to keep it as consistent as you can.

The same is true with the El Prez. The guy who always shoots it hero-or-zero is going to have times when he's awesome and times when he's a disaster. They guy who tries to make tiny incremental improvements won't have "bad" runs nearly as often but also won't have those "flashes of brilliance" that break through plateaus.

Assuming our goal is about on-demand one time performance (e.g., a fight) as opposed to simply a better Mean (e.g., a 12-stage match) then I think practice needs to involve both factors at various times. You need to build consistency so that even your outlier "bad" performance is still acceptable but you also need to push your envelope at times so you continue to move your Mean toward better overall performance.

Something to think about: how does USPSA scoring (specifically, stage points, or comparative scoring) impact hero-or-zero vs. consistency compared to IDPA scoring (absolute scoring)?

[ToddG]

As far as positive vs negative feedback (which wasn't really touched on in any meaningful way in the video) as it relates to shooting, I've always thought it would be interesting to put a canine shock collar around someone's leg and zap them whenever they made a miss or fumbled a draw/reload or failed to meet a time standard during a drill. :cool:

More realistically, I do think we see some negative feedback training in shooting and the results are worth considering.

The most obvious is shooting steel. A lot of folks talk about it in terms of positive feedback but really it's both: you know if you hit (positive) or missed (negative). At a certain point in your skill development, you expect hits so the positive effect is probably pretty minor while the negative effect of a miss is greater.

Calling your shots is another great instant positive/negative feedback tool. I think it's one reason why people who learn to call their shots effectively seem to hit a steep improvement curve... they're getting feedback instantly on every shot and that's helping them program "right" into their eyes and fingers faster and with fewer false positives.

The antithesis is the guy who fires 500 rounds at a single paper target and never tapes or otherwise indicates which shots have been fired previously. He has absolutely no idea whether the rounds he's firing are hits or misses. He has no idea whether he's doing well or screwing up. I took a class once in which students shot multiple Bill Drills on the same target at multiple distances without ever taping the targets between runs. Then we just counted hits. If you weren't very good at shot calling, you had no idea whether you'd got all your hits at 5yd and none at 25yd or something in between. It was borderline useless practice.

[Mr_White]

Something to think about: how does USPSA scoring (specifically, stage points, or comparative scoring) impact hero-or-zero vs. consistency compared to IDPA scoring (absolute scoring)?

It was interesting to me to hear the assertion a while back that successful 'match pace' in USPSA needed to be not pushing at all, but simply performing at the level you can execute very reliably. I'd say I've found that to be true for myself. Hero-or-zeroing stages in general does not seem to usually lead to winning a USPSA match overall. There may be a couple of stage wins, but that disaster stage where you are in 32nd place tanks your overall results. Ask me how I know. : ) But, a lot of times in USPSA, finishing between 2nd and 4th on every stage with no disaster adds up to 1st place overall.

Classifier stages seem oriented toward zero-or-heroing more than stages overall, due to the filtering applied in the classification calculations. The other day I was thinking of how radically different I suspect people would shoot classifiers if every single score counted toward your classification, even if you zeroed the stage, and you would go down in classification if your percentage warranted it. I think it would be B-class city up in that classifier stage. Maybe I am wrong, but that's what it seems like to me.


Calling your shots is another great instant positive/negative feedback tool. I think it's one reason why people who learn to call their shots effectively seem to hit a steep improvement curve... they're getting feedback instantly on every shot and that's helping them program "right" into their eyes and fingers faster and with fewer false positives.

Great point. That is truly one the greatest gateway benefits of better shot calling - you can self-diagnose since you can see what the gun is doing. That really does open the door for accurately-targeted self-improvement.

[justintime]

Through coaching people to be faster behind the wheel I have noticed it totally depends on the persons personal mindset. When coaching younger kids who are new, if they are trying to decide if they even like the activity then positive reinforcement is paramount.

Likewise men who are just trying to learn how to drive safely and get a feel for something, or people who have false senses of reality usually do very bad with negative reinforcement. These people either don't want to learn they are making mistakes, or never cared to truly get get better in the first place if it were to put their ego at risk. For some people as odd as it sounds, training is an exhibition of their skills and others it's a tool to survive doing something cool and learning is not actually their end goal. I actually ask everyone what their goals are before hand so I know how to handle them.

Likewise I have had people come to me with drive and ambition who want to win or be the best. These people sometimes will get down after negative reinforcement but instead of quitting they bounce back and get better. I think you have to use positive and negative in this aspect as it is important for people to not get a sense of false reality if they are expecting to truly be good, likewise they also need encouragement when they do something right. It is totally individually based on commitment if a person can handle both - and those people are always honest with themselves and end up being my best students.

You can see the two different kinds of people on forums too. People post pictures or videos to be evaluated and based off of the feedback some people freak out and some people say thanks and make changes if necessary. I've noted the people who ask for help who then criticize the advice are just exhibition students and not real learners.

[TheTrevor]

I've always thought it would be interesting to put a canine shock collar around someone's leg and zap them whenever they made a miss or fumbled a draw/reload or failed to meet a time standard during a drill. :cool:

I would pay cash money to watch people training like this. Especially some of the jokers I watched at the range yesterday. Double if I could be the one with the remote control.

[nycnoob]

Before I got into shooting I took a bunch of classes about positive reinforcement,
from the experts, the animal trainers who do it.

I found it very interesting and it does help to explain some behaviors
that the MBA types really get wrong in a big way.

The big expert in the field is Karen Pryor
http://www.clickertraining.com/

and her famous book "Don't Shoot the Dog"
http://store.clickertraining.com/dont-shoot-the-dog-karen-pryor.html

She regularly runs three day workshops filled with lectures and exercises
(some people oriented and some animal oriented, often the venue allows dogs
and one of her instructors has a seeing eye, miniature horse)
http://www.clickertraining.com/clickerexpo?source=eventsnav


Many of her instructors are gymnastic coaches as well. They adapted the principles to teaching
teen age girls gymnastics
http://www.tagteach.com/


I devoted a whole bunch of time to this side educational project so I could speak on it for a long time
but this is a gun board, so I will mostly leave you with the links and get back to shooting.

Oh but the one interesting thing: they are very big on positive reinforcement and believe that it is
the best (for a host of reasons) way to teach most everything, however I do not think they
can train guard dogs, which would be of interest to many members of this board.