RE: [Heartlogic-dev] new idea

[William L. Jarrold]

I hope you don't mind me putting this back on heartlogic-dev.

On Wed, 4 May 2005, Joshua N Pritikin wrote:

> On Tue, 2005-05-03 at 23:41 -0500, William L. Jarrold wrote:
> > > > "Vienna is wet."
> > No I'm not joking.  This is a simple study, using your very good
> > hot-or-not type interface code it should be easy to put together in an
> > hour or so.  I want about 5-10 judgements per item.  It is a
> > constrained study.  If you get the one up about Vienna above, then we
> > can tune it and get five up.  Then start announcing it a little.
>
> Oh!  Well .. why not?

Great.

> Can you give me a list of items like "Vienna is wet." ?

They were in my previous long email about rumination.  I have copied them
here...

(k) "Jesus loves his father bc Jesus is God's son."

(l)  "The intensity of contempt that Hillary Clinton feels for Bill Clinton
was zero.  This is because ...

(a) Hillary is married to Bill and 
(b) spouses love each other and 
(c) love is the opposite of contempt and

(d) if you feel emotionx and y is the opposite emotion of x then you
feel zero amount of y.

Hillary feels zero amount of conempt for Bill Clinton

(j) "Vienna is wet."
(e) "Rivers are a kind of water."
(f) "If water touches x then x is wet."
(g) "The Danube is a river."
(h) "The Danube runs through Vienna."
(i) "If a river runs through a region is touches that region."

...I say lets just toss 'em up there and then start improving things.

> I assume you
> are just making these up?

No.  Making them up would be absurd.  My mind reading system is
starting to infer that you did not read my rumination email very
carefully nor my AAAI paper about this topic.

> Or are you getting them from a real KR
> rumination?

Yes, yes yes.

Here are some quotes from my rumination email...

Cyc concluded that (j) "Vienna is wet."

Cyc concluded many
things.  The huge majority were true.  Some were vacously true.  But,
even 1 percent false deduction can mean we can not rely on the thing.
Some conclussions were fun or cute, like...

(k) "Jesus loves his father bc Jesus is God's son."

...some were kind of funny...like

(l)  "The intensity of contempt that Hillary Clinton feels for Bill Clinton
was zero.  This is because ...

> Is this study anonymous or do you want to require login?

Anonymous is fine.

As Josh pointed out, it would be nice to be able to catch malicious
people who try to mess up the ratings.  So, in time we will want to be
able to track identity.

> I assume you want a rating and an explanation?  Or just a rating?

Right now, I don't care much either way....Actually, after
articulating the below, I concluded that while we pilot this we should
ask for an explanation.

Maybe Josh or you have a compelling opinion.  Here's my thought process...

Argumnets against explanation:

(a) An explanation slows people down.

(b) Given how bad I have been about pouring over already collected
data, why should we pester participants with writing explanations that
I will ignore.

Argumnets for explanation:

(a) explanations also helps us understand their thought process and
helps us explain weird answers.

(b) when we are piloting this, we especially want an explanation.

...in sum, the above argues lets do the pilot with an explanation.

Once we get this cleaned up then we'll think about whethher we want
explanation of not for real.

> > Well, there is a hidden issue here.  Part of it is my selfishness.  I want
> > to work on my pet projects.
> > ...I have been wanting to collect data on the "vienna is wet" type
> > stuff since before I wrote that paper.
>
> There are a few practical issues.
>
> 1. What about the replication of your dissertation?

It should and will happen.  But it seems to me a study in obviousness.
The best rationale for it is this: If we want to publish www collected
data, being able to say "Elsewhere we replicated a study that was done
with real live human subjects."  Therfore, the data you are seeing in
this study is probably valid.

Well, there are many factors in doing a good study.  Being able to
say, "Hey, we did this other study (i.e. a replication of Jarrold's
dissertation) and it came okay, therefore, weenies filling out surveys
on the web produce usable data." helps our case BUT is not necessary.
In fact, people like the Open Mind people (and maybe a guy at ISI who
has done a similar gather-data-from the web) have made interesting
waves in the AI world and have not done the intense level of checking
WWW participants data against "real human, college sophmore" type
data....Thus, the dissertation replication will help us obtain
confidence in our data but it is not a requirement.  In this way the
replication is like icing on the cake.

What *is* a requirment for our new studies is something up there for
participants to fill out else besides my dissertation!  Better to work
on the necessary stuff BEFORE the icing on the cake.

> 2. We need a server that is better situated than my home machine.  1and1
> is not suitable.  http://www.linux-vserver.org/ looks pretty cheap.
>
> > And another part of it is fear of taking on
> > something that is very hard to figure out, get a handle on...The recent
> > studies you have proposed have been very hard for me to wrap my mind
> > around.
>
> OK, but I hope this is a temporary condition due to job pressure, etc.

I think job pressure etc will always be there until I get this work
directly funded or I magically become independently wealthy.  Either
will be years from now.

Still, I can read your emails, think about your studies.  It will just
be slow going, high risk.

> > It will take me many many hours to get my mind around your study and
> > make it methodologically sound to the point where I would feel
> > comfortable putting my name behind it.  Finding the time to work on
> > this is so precious that I want to give highest priority to things
> > that have a high chance of success for minimal effort.
>
> Fair enough.
>
> Do you have any objection with running my goal-type pilot study in
> rotation since all the code is written already?

No objections.  Go for it.  I will try to reply to the emails that you
and Josh went around about.  Here's what might be a helpful hint: try
to make your language as clear to an ignorant person as possible.
Also, minimize the amount of text on each item.

BUT, what we should not do is publicize any study until we all agree
that it is up to snuff to risk "blowing our wad" on treasured
participants.

Bill

> --
> If you are an American then support http://fairtax.org
> (Permanently replace 50,000+ pages of tax law with about 200 pages.)