> -----Original Message-----
> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On
> Behalf Of Chad Cooper
[snip]
> "What about making an assumption that if a respondent who rarely
> responds to
> messages, does respond to someone who posts regularly, that this could be
> considered a meme that had an unusual effect. This could then be
> represented
> as a vector.
Got it -- Bayesian probability. Another way of saying it is that common
events are insignificant, so the fact that a person who replies to every
thread has replied to a given thread is not significant. "Rarely" is the
key. I've used the 80/20 rule of thumb on this kind of thing with decent
results. In other words, when someone whose posting frequency is around the
20th percentile posts, that may be significant. However, I haven't seen
much reason to have faith in the significance of such analysis by itself.
> The converse of this is that active responders are less 'vunerable' to
> individual meme effects. For example, I personally am more likely
> to respond
> to a meme, than to issue a new one to the group. It is also more
> likely that
> I will get little if any response back from the group, when compared to
> those people who post multple messages a day. I think that there is a
> relationship where if a person posts a lot of messages, that
> person is much
> more likely to get someone else to respond back to a individual
> or specific
> message. This suggests that there is a vunerability to those people who do
> not post often.
Yep -- the seemingly wacko posting from a stranger is much more likely to be
ignored v. the same thing from someone who is well known. That's why
reputation (and more specifically, influence) matters and is an important
starting point for metrics.
> The degree of effect possible from any specific meme on any one individual
> is directly related to the sum of memes presented to the group from that
> same individual. Summarized, Memes are cumulatively defensive. The best
> example would be John and Jeroen, two of the top posters. They commonly
> appear to be 'immune' to each others ideas, and spend great
> amount of effort
> to defend their existing ideas... going for penetration into the other's
> thick skulls - The kill shot.. hitting a nerve... etc.
> I would think that this could be mathematically described much
> better, but I
> ain't no genius."
You mean that people who talk a lot are less likely to be taken seriously?
Although there's surely some truth in that statement, I doubt (and hope!)
that it's not a linear relationship. Many of these kinds of things involve
thresholds which, when crossed, result in non-linear changes. Gladwell's
"The Tipping Point" is a terrific little book about such phenomena. It's
somewhat related to notions of chaos and complexity, showing how small
changes can have dramatically large effects -- once the tipping point is
reached.
I wonder if I could come up with an average "tipping point" in terms of
posting frequency, beyond which people are rarely taken seriously. It's
probably related to overall traffic, I suppose. Some lists have so much
traffic that nobody manages to dominate, I think. Some have so little that
no one could.
> So I guess what I am saying is that "Meme conversion" is a number value
> (scalar for you adults!) that represents a meme having an effect upon
> another. 'Meme immunity" is the degree that one resists new ideas and
> cummulatively defends existing ideas. I don't know if this would be scalar
> or boolean.
You're thinking that way because of tipping points. It looks scalar at many
points along the way, but the dramatic non-linear change looks like a
Boolean, though it's probably not useful to think of it as one. More like a
quantum shift, so to speak.
> Also remember that I have nothing to really base this hypothesis on other
> than gut feeling. However it can be tested using the scientific
> method using
> the vote tally method I described.
If people gave honest feedback... which I doubt. I would have a hard time
filling out a form that indicated that I don't take person X seriously,
since I think of myself as open-minded. Sort of like asking people what
kind of television they watch -- PBS comes out much higher than actual
ratings indicate. Or asking people if they are above-average drivers -- 80
percent say they are.
> You could use the breaks where people insert their responses in
> messages. If
> there is a break in the >, that would delimit an idea. This poses
> a problem
> if someone does not delimit their response. If a common interface
> is used it
> would prevent these problems. At worst, a vote marker would be put in an
> inappropiate place. You could separate out those vote markers that get no
> response as errors.
Oh. Hmm. I'm not sure I even want to try to identify different ideas
within a given message. The whole notion of capturing a meme is
problematic, although feature vectors head in that direction. The problem
is that there's often valence involved (positive, negative opinions, for
example), which is incredibly difficult to parse out of natural language.
We often use words for exactly opposite meanings. Since that statement
often confuses people, here's an example.
Draw me a model of that car.
I just bought the new model of the Thunderbird.
In the first case, "model" does not mean a real car, while in the second, it
does. This kind of usage is plain as day to us human beings, but confusing
as heck to software, which gets bogged down very, very fast trying to figure
out the context.
Nick
