final poll results.

[John Anderson]

Content-Type: text/plain; charset="us-ascii" ; format="flowed"

Seven more respondents answered after my initial distribution of the 
poll results on Nov 6, raising our total to 19.  There still are some 
significant non-respondents (you know who you are) but this probably 
gives a pretty good picture of where things stood as of a couple of 
months ago (the 7 responded all shortly after the partial results 
were distributed).   The conclusions are not dramatically different. 
The general conclusions seem that 5.0 is a keeper with lots of use 
already, that buffers are a big win, that the jury is out on just how 
to treat the goal stack, that there are problems with activation/ 
competitive latency but few are experiencing them yet, and that 
production compilation is receiving little use so far.

>1. With respect to 4.0:
>a. Have you ever worked with it?
  I know there is at least one person who has only worked with 5.0 but 
all 19 of these respondents had worked with 4.0

>b. How many research models would you estimate that you have produced in 4.0?
The median was between 5 and 10.

>c. Are you currently using it for research?
9 respondents are

>d. Have you used it (or do you intend to use it) for teaching?
10

>2. With respect to 5.0
>a. Have you every worked with it?
16 .

>b. How many research models would you estimate that you have produced in 5.0?
This is best answered by the actual numbers -- 5 "0"'s, 7 "1"'s, 4 
"2"'s, 2 some, and 1 "4".

>c. Are you currently using it for research?
15

>d. Have you used it (or do you intend to use it) for teaching?
13

>3. With respect to the principal features on which ACT-R 5.0 differs 
>from 4.0, how much experience have you had with
>a. The buffer control structure
Roughly, 6 "lots", 8 "some", 3 "little", 2 "none".

>b. The downgrading of the goal stack
This response could better be characterized binary with 14 having 
experienced it and 5 not.

>c. The competitive latency equation and downgrading of the strengths 
>of association
1 "lots", 5 "some", 4 "little", and 9 "none"

>d. Production compilation
2 "lots", 2 "some",  15 "none"

>4. With respect to the principal features on which ACT-R 5.0 differs 
>from 4.0, how would you evaluate
>a. The buffer control structure
16 "good-to-great", 3 qualified.  The one suggestion that was made 
was that we needed greater consistency in our use of buffers.

>b. The downgrading of the goal stack
7 "good-to-great", 10 "problematic", and 2 no response.  None of the 
problematic seemed to want to go back to 4.0 but they still had their 
problems.  Two complained about returning values, one about the time 
to retrieve goals, one about the lack of a pop-on-failure, one about 
goal forgetting, and one just the struggle in adjusting to the new 
style.

>c. The competitive latency equation and downgrading of the strengths 
>of association
2 OK, 13 had no comment, 4 noted problems -- for one that was lack of 
an associative learning mechanism; for the other 2 it was the large 
effects of number of competitors.

>d. Production compilation
1 "good", 7 "qualified good", and 11 no comment.  The qualifications 
largely involved doubts about the mechanisms for gradually 
introducing the production rules.  I know the production compilation 
mechanism remains a little brittle but there has been too little use 
to produce any annoyance with that yet and it is being at least 
somewhat improved through the experience of the few current users.

>
>5. Any other comments?
One person felt that they needed simultaneous access to more than the 
two chunks that can be held in goal and retrieval buffer.

One person wanted more support in terms of materials for teaching 
5.0.  Another person wanted more documentation along the lines of the 
Atomic Components of Thought.

One person wanted a failure chunk in the visual buffer if there was 
no object at the attended location. That same person wanted 
variability in cycle time.

One person wanted a representation of time, haptic feedback/learning, 
an ability of the hand to do things like move a throttle, and worried 
that the real window interface in 5.0 is compatible with real windows 
in ACL 6.0.

One person wanted more outlets for publishing  ACT-R models. 
However, I think as a whole the user community has been quite 
successful at getting their models published.

One person felt that there were problems satisfying style constraints 
on chunk creation in either 4.0 or 5.0.

Another person omplained about the "state" or "step" slots that tend 
to be part of goals in 5.0 models.

One person complained about the extra burden in dealing with the 
environment in 5.0.

That same person felt that there needed to be better debugging 
features for 5.0 particularly to deal with the parallelism.
-- 

==========================================================

John R. Anderson
Carnegie Mellon University
Pittsburgh, PA 15213

Phone: 412-268-2788
Fax:     412-268-2844
email: ja at cmu.edu
URL:  http://act.psy.cmu.edu/
--============_-1201617252==_ma============
Content-Type: text/html; charset="us-ascii"

<!doctype html public "-//W3C//DTD W3 HTML//EN">
<html><head><style type="text/css"><!--
blockquote, dl, ul, ol, li { margin-top: 0 ; margin-bottom: 0 }
 --></style><title>final poll results.</title></head><body>
<div><font face="New York" color="#000000">Seven more respondents
answered after my initial distribution of the poll results on Nov 6,
raising our total to 19.  There still are some significant
non-respondents (you know who you are) but this probably gives</font>
a pretty good picture of where things stood as of a couple of months
ago (the 7 responded all shortly after the partial results were
distributed).   The conclusions are not dramatically
different.  The general conclusions seem that 5.0 is a keeper
with lots of use already, that buffers are a big win, that the jury is
out on just how to treat the goal stack, that there are problems with
activation/ competitive latency but few are experiencing them yet, and
that production compilation is receiving little use so far.</div>
<div><br></div>
<blockquote type="cite" cite><font face="New York" color="#000000">1.
With respect to 4.0:</font></blockquote>
<blockquote type="cite" cite><font face="New York" color="#000000">a.
Have you ever worked with it?</font></blockquote>
<div><font face="New York" color="#000000"> I know there is at
least one person who has only worked with 5.0 but all 19 of these
respondents had worked</font> with 4.0</div>
<div><br></div>
<blockquote type="cite" cite><font face="New York" color="#000000">b.
How many research models would you estimate that you have produced in
4.0?</font></blockquote>
<div>The median was between 5 and 10.</div>
<div><br></div>
<blockquote type="cite" cite><font face="New York" color="#000000">c.
Are you currently using it for research?</font></blockquote>
<div>9 respondents are</div>
<div><br></div>
<blockquote type="cite" cite><font face="New York" color="#000000">d.
Have you used it (or do you intend to use it) for
teaching?</font></blockquote>
<div>10</div>
<div><br></div>
<blockquote type="cite" cite><font face="New York" color="#000000">2.
With respect to 5.0</font></blockquote>
<blockquote type="cite" cite><font face="New York" color="#000000">a.
Have you every worked with it?</font></blockquote>
<div>16 .</div>
<div><br></div>
<blockquote type="cite" cite><font face="New York" color="#000000">b.
How many research models would you estimate that you have produced in
5.0?</font></blockquote>
<div>This is best answered by the actual numbers -- 5 "0"'s,
7 "1"'s, 4 "2"'s, 2 some, and 1
"4".</div>
<div><br></div>
<blockquote type="cite" cite><font face="New York" color="#000000">c.
Are you currently using it for research?</font></blockquote>
<div>15</div>
<div><br></div>
<blockquote type="cite" cite><font face="New York" color="#000000">d.
Have you used it (or do you intend to use it) for
teaching?</font></blockquote>
<div>13</div>
<div><br></div>
<blockquote type="cite" cite><font face="New York" color="#000000">3.
With respect to the principal features on which ACT-R 5.0 differs from
4.0, how much experience have you had with</font></blockquote>
<blockquote type="cite" cite><font face="New York" color="#000000">a.
The buffer control structure</font></blockquote>
<div>Roughly, 6 "lots", 8 "some", 3
"little", 2 "none".</div>
<div><br></div>
<blockquote type="cite" cite><font face="New York" color="#000000">b.
The downgrading of the goal stack</font></blockquote>
<div>This response could better be characterized binary with 14 having
experienced it and 5 not.</div>
<div><br></div>
<blockquote type="cite" cite><font face="New York" color="#000000">c.
The competitive latency equation and downgrading of the strengths of
association</font></blockquote>
<div>1 "lots", 5 "some", 4 "little", and
9 "none"</div>
<div><br></div>
<blockquote type="cite" cite><font face="New York" color="#000000">d.
Production compilation</font></blockquote>
<div>2 "lots", 2 "some",  15
"none"</div>
<div><br></div>
<blockquote type="cite" cite><font face="New York" color="#000000">4.
With respect to the principal features on which ACT-R 5.0 differs from
4.0, how would you evaluate</font></blockquote>
<blockquote type="cite" cite><font face="New York" color="#000000">a.
The buffer control structure</font></blockquote>
<div>16 "good-to-great", 3 qualified.  The one
suggestion that was made was that we needed greater consistency in our
use of buffers.</div>
<div><br></div>
<blockquote type="cite" cite><font face="New York" color="#000000">b.
The downgrading of the goal stack</font></blockquote>
<div>7 "good-to-great", 10 "problematic", and 2 no
response.  None of the problematic seemed to want to go back to
4.0 but they still had their problems.  Two complained about
returning values, one about the time to retrieve goals, one about the
lack of a pop-on-failure, one about goal forgetting, and one just the
struggle in adjusting to the new style.</div>
<div><br></div>
<blockquote type="cite" cite><font face="New York" color="#000000">c.
The competitive latency equation and downgrading of the strengths of
association</font></blockquote>
<div>2 OK, 13 had no comment, 4 noted problems -- for one that was
lack of an associative learning mechanism; for the other 2 it was the
large effects of number of competitors.</div>
<div><br></div>
<blockquote type="cite" cite><font face="New York" color="#000000">d.
Production compilation</font></blockquote>
<div>1 "good", 7 "qualified good", and 11 no
comment.  The qualifications largely involved doubts about the
mechanisms for gradually introducing the production rules.  I
know the production compilation mechanism remains a little brittle but
there has been too little use to produce any annoyance with that yet
and it is being at least somewhat improved through the experience of
the few current users.</div>
<div><br></div>
<blockquote type="cite" cite><font face="New York"
color="#000000"><br></font></blockquote>
<blockquote type="cite" cite><font face="New York" color="#000000">5.
Any other comments?</font></blockquote>
<div><font face="New York" color="#000000">One person felt that they
needed simultaneous access to more than the two chunks that can be
held in goal and retrieval buffer.</font></div>
<div><font face="New York" color="#000000"><br></font></div>
<div><font face="New York" color="#000000">One person wanted more
support in terms of materials for teaching 5.0.</font>  Another
person wanted more documentation along the lines of the Atomic
Components of Thought.</div>
<div><font face="New York" color="#000000"><br></font></div>
<div><font face="New York" color="#000000">One person wanted a failure
chunk in the visual buffer if there was no object at the attended
location. That same person wanted variability in cycle
time.</font></div>
<div><font face="New York" color="#000000"><br></font></div>
<div><font face="New York" color="#000000">One person wanted a
representation of time, haptic feedback/learning, an ability of the
hand to do things like move a throttle, and worried that the real
window interface in 5.0 is compatible with real windows in ACL
6.0.</font></div>
<div><font face="New York" color="#000000"><br></font></div>
<div><font face="New York" color="#000000">One person wanted more
outlets for publishing  ACT-R models.  However, I think as a
whole the user community has been quite successful at getting their
models published.</font></div>
<div><br></div>
<div>One person felt that there were problems satisfying style
constraints on chunk creation in either 4.0 or 5.0.</div>
<div><br></div>
<div>Another person omplained about the "state" or
"step" slots that tend to be part of goals in 5.0
models.</div>
<div><br></div>
<div>One person complained about the extra burden in dealing with the
environment in 5.0.</div>
<div><br></div>
<div>That same person felt that there needed to be better debugging
features for 5.0 particularly to deal with the parallelism.</div>

<div><font color="#000000">-- <br>
</font></div>
<div><font
color="#000000"
>==========================================================</font></div
>
<div><font color="#000000"><br></font></div>
<div><font color="#000000">John R. Anderson<br>
Carnegie Mellon University<br>
Pittsburgh, PA 15213<br>
<br>
Phone: 412-268-2788<br>
Fax:     412-268-2844<br>
email: ja at cmu.edu<br>
URL:  http://act.psy.cmu.edu/</font></div>
</body>
</html>
--============_-1201617252==_ma============--