Posted by Klavot on July 5, 2007, at 3:11:14
In reply to Re: blocks, posted by Dr. Bob on July 4, 2007, at 16:35:34
> > there are moments when one impulsively does things, and then is taken up short by what one has done.
> >
> > If it's a question of giving people a chance to repeat an experience, so as to learn enough from the repetitions that it would actually help change behavior, allowing people to come back and make mistakes and being blocked more often would, for me, make the learning process more probable.
> >
> > Honore
>
> I agree, it can be hard if someone's impulsive or needs many repetitions to learn, but if people can't make mistakes as often, there's less overall incivility and more supportive overall atmosphere.
>
> --
>
> > What about r = 39.
> >
> > It is the smallest r that guarantees that after 6 months, previous incivility will still factor. So even if the previous block was only one week, and that expired 6 months ago, then
> >
> > B = S + D*[3^(-P/39)] = S + 1*[3^(-24/39)] = S + 0.51 -> S + 1 week.
> >
> > Klavot
>
> Thanks for working on this. Hmm, powers of e probably are about as intuitive than powers of 3, and if we set:
>
> 0.5 = exp(-24/r)
>
> then:
>
> r = 24/ln2 ~ 35Yes, when working with exp(x) you would need r = 35. But for 3^x you get r = 39. But we can stick to exp(x) if you are interested, that's fine by me. My concern was simply that the calculator that comes bundled with MS Windows does not seem to have a function for retrieving e. This means some posters would have no way of calculating block lengths or verifying the arithmetic behind block lengths, which might cause complaints.
>
> Another issue is when a poster is immediately uncivil again. Currently, if it's impersonal and S = 2:
>
> B = S * (D - P/r) = 2 * D
>
> But with your formula and S = 1:
>
> B = S + D * exp(-P/r) = 1 + D
>
> It makes the formula more complicated, but to add the exponential decay and keep the current doubling (exponential growth), it could be:
>
> B = 1 + (SD - 1) * exp(-P/r)
>
> with S = 2 or 3. So if D = 48 and P = 147 and it were impersonal, we would still get:
>
> B = 1 + (2 * 48 - 1) * exp(-147/35) = 2.42 -> 2 weeks
>
> In Zenhussy's case, however, it was personal, not impersonal, so it would've been an extra week:
>
> B = 1 + (3 * 48 - 1) * exp(-147/35) = 3.14 -> 3 weeks
>
> BobOK. My thinking was that with a formula of the form
B = 1 + (SD - 1) * exp(-P/r),
you have a situation where the severity of a current incivility is extrapolated to previous incivilities, which is why I would prefer not to multiply the term D*exp(-P/r) with S. However, if this is the direction you want to go, then what about the original formula that I posted lower down in another thread:
B = S(D*exp(-P/r) + 1).
This formula allows for doubling or tripling as you see fit, while also being applicable to both repeat incivilities as well as first incivility. For a first incivility, we get
B = S(0*exp(-0/r) + 1) = S.
For the other formula, we get
B = 1 + (SD - 1) * exp(-P/r) = 1 + (S*0 - 1)*exp(-0/r) = 0,
which means you would have to split the formula into cases for a first offense versus repeat offense.
But all this is just a gimmick. The basic idea is to introduce some kind of exponential decay for previous blocks. After that one can build a formula to fit your desired blocking model.
Klavot
poster:Klavot
thread:762973
URL: http://www.dr-bob.org/babble/admin/20070702/msgs/767791.html