[FRIAM] Screening off

[Eric Charles]

If 1 co-occurs with 2 50% of the time, and 2 co-occurs with 3 50% of the
time, and that 1 never occurs with 3 except when 2 has occurred, then:
* Given 1 happens, then you would guess that 3 will happen 25% of the time.
* But if you knew whether or not 1 had caused 2 in a particular instance,
you could significantly improve that guess (shifting to 0% or 50%).
* In contrast, if you knew 2 happened, you would guess that 3 would happen
50% of the time, AND it wouldn't change anything if you found out whether
or not 1 had happened.
<echarles at american.edu>

Now, in place of "co-occurs" some people want to substitute "cause", which
is frustrating and confusing to you, because you are an experimentalist.
This is an "observational" approach to trying to draw conclusions, not a
"true experiment" approach. At no point do you get to manipulate the
variables and see what happens. You *just *look at the data that was
collected, and conclude A) that 2 causes 3, and B) that to the extent that
1 causes 3, it causes 3 via 2, and in no other fashion. And, as an
experimentalist, you know that's a pretty non-ideal approach to getting at
causation, no matter how schmanzy the algorithms get.

In contrast, let us imagine we are studying the vocalizations of macaques,
and we notice that a certain vocalization happens, which haven't heard very
often, seems to co-occur with the presence of leopards. We think "Hey,
maybe seeing a leopard is causing the cries" ---> Presence of leopard
causes seeing of leopard, seeing of leopard causes cries. In this case, we
would find that a much messier relationship. It would look something like
this:
* Presence of a leopard co-occurs with seeing a leopard 80% of the time.
* Seeing a leopard co-occurs with cries 80% of the time.
* IF seeing the leopard screened off, then when a leopard is present, you
would expect a given macaque to make a cry 64% of the time - if, and only
if, they see the leopard.
* But you observe a much higher rate of response when you do your
observations. What's happening? Why is the leopard-call response not
"screened off" by having seen the leopard?
* Well.... there are OTHER ways that the presence of a Leopard causes
cries, such as hearing a leopard, or smelling the leopard. (Plus leopard
calls are mimicked.) Because of that, it wouldn't be surprising to have
leopard calls occur 80 or 90% of the time a leopard is nearby.



On Sat, Dec 4, 2021 at 7:38 PM Nicholas Thompson <thompnickson2 at gmail.com>
wrote:

> Frank,
>
> Still need help. Given events 1, 2, and 3, 3 has been screen off by 2 from
> 1, if  the probability that 3 occurs given that 2 has occurred is equal to
> the probability that 3 occurs given that both 2 and one have occurred.
>  As I understand mathematics this equality requires that the probability of
> 1 occurring is 1.00.  Another way to say that is that the probability that
> 3 occurs  if 2 has occurred is the same as the probability that 3 has
> occurred if 2 has occurred, and 1 has already occurred.  What's the fun in
> that?  In other words, given the possibility of other causes for 2, the
> fact that 2 occurs gives us relatively little evidence that 1 has
> occurred.  Isn"t this true of all causal abduction?
>
> N
>
> .-- .- -. - / .- -.-. - .. --- -. ..--.. / -.-. --- -. .--- ..- --. .- - .
> FRIAM Applied Complexity Group listserv
> Zoom Fridays 9:30a-12p Mtn UTC-6  bit.ly/virtualfriam
> un/subscribe http://redfish.com/mailman/listinfo/friam_redfish.com
> FRIAM-COMIC http://friam-comic.blogspot.com/
> archives:
>  5/2017 thru present https://redfish.com/pipermail/friam_redfish.com/
>  1/2003 thru 6/2021  http://friam.383.s1.nabble.com/
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://redfish.com/pipermail/friam_redfish.com/attachments/20211205/a9959209/attachment.html>