Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[wierd]

Oh, the research would certainly be useful for predicting the behavioral trends of students, for sure.

The behavioral sciences department would eat it up like candy.  (Simply because the model does not produce sentience in a machine, does not mean the model is without purpose.)

[Trekkin]

Oh, the research would certainly be useful for predicting the behavioral trends of students, for sure.

The behavioral sciences department would eat it up like candy.  (Simply because the model does not produce sentience in a machine, does not mean the model is without purpose.)

Aww crud, really? I guess I'd better stop throwing protein folds at our AIs and expecting them to spontaneously write me a sonnet about the meaning of life, then. /s

Seriously, though, there's no proposed model here. All we have is a suggestion that we load data into an AI in order to learn about it. My point was that AI doesn't operate with the transparency necessary for that to be a productive line of inquiry.

EDIT: PCA could, maybe?

[wierd]

You would need to quantize and pre-process that data to feed it meaningfully to an AI. (otherwise the lack of structure would make the data essentially into noise.)  That structured data could then be used to make said models, that the behavioral sciences people would eat up like candy.

(the most immediate and simplistic preprocessing would be temporal-spacial correlation of excitation states)

[Trekkin]

That structured data could then be used to make said models

This is the definition of zeteticism, not empiricism. This is literally not science. Thinking this is how science works is how Flat Earthers happen.

[wierd]

Science is about testing observations.  This implies a temporal component, otherwise there is no causality, and there is no means of determination.


A computer has no inherent sense of time, or order, for data.   Millions of individual samples, devoid of any context, either in time, or in space, are impossible to derive much meaning from.  It is completely without structure. 


EG-  "If I hit this ball, it will move in the direction in which I swung my club" is an observation that can be tested.

"I swing my club. Ball moves."  and "Ball moves. I swing my club."  with no data on direction, or order, it is impossible to derive which is causal.


That is what I mean when I say "structured."  The structure is not arbitrarily selected (which is just nonsense.)  It is an active and necessary component in the data collected.

EG-- instead of "We collected 10*e^15 excitations during the experiment." or "We collected 200 million excitations in the prefrontal cortex during the experiment."  you have "Between this timestamp and this timestamp, we collected 50 million unique excitation series originating in the prefrontal cortex, which propogated at X rate to the hypothalamus."

Which gives you a time, and a location, from which to map events in the environment, and produce models.

[Trekkin]

Okay, wierd. There's a lot to unpack there, but let's start with the fundamentals:

What, in your own words, is zeteticism?

[wierd]

Zetectism:

"Does the ball go in the direction of my club when I strike it?"
 (Question asked first, then experiment.)

"When I strike this ball with my club, it goes in the direction I struck it.-- True/False"
(Stated theory formulated from observation, tested for falsity.-- Normal science.)


In this case-

"At timestamp X through Y, N Series unique excitations with a specific identifying distribution were recorded propogating from the prefrontal cortex to the hypothalamus."

It is raw data.  It is useful for formulating a hypothesis that can then be tested.  EG, "The time between X and Y is lunch time." can be added to create a theory "The data being generated and sent to the hypothalamus is the mental state caused by the flavor data. (statement)."  This can then be tested-- Participant skips lunch, new series produced.  Evaluate differences in series.

If inconclusive-- Repeat prior experiment while assuring subject is not hungry (say, introvenous glucose administered, etc).  Collect series, examine for differences in patterns over timestamp sampled.


You still need the timestamp, and the localities in which the series originate or migrate to.  Otherwise you cannot produce a cogent theory for the excitation pattern.

[Trekkin]

Not quite, no. Zeteticism is literally this:

That [...] data could then be used to make [...] models

There's no questions involved at all, just taking a bunch of data and building a mechanism out of it. To use your ball example, it's seeing that you can hit the ball in a certain direction with your club and deciding why that happens based solely on the congruence between your theory and the data you used to make it. Alternatively, it's seeing that the world looks flat to you and deciding that it is. That's why it is awful science: you can make any theory you want right down to "God created ad hoc physics to produce exactly these data" and have it accord with 100% of observation, so there's no predictive value.

Empiricism doesn't let you do that. You build a model (which is really just a mechanism with defined input and output parameter sets) based on theory and then identify a data set that could falsify it. Then you go do the experiments necessary to get that data and either disprove or fail to disprove the most relevant null hypothesis or equivalent. By the way, this:
 

Science is about testing observations. 

Is also wrong for this reason. We don't test observations. We test predictions.

There's also a more minor point to be made that those predictions are not always about causes. Predictions are a logical expression of a pattern, and patterns can exist  without a time component. It's totally valid to predict that if set X has Y traits it will also have Z traits because of some mechanism and go off and go look for an X that has Y but not Z or otherwise try to disprove that assertion.  Y and Z can even have spacelike separation and the process still works. Sure, there's implicitly a cause somewhere, but mechanisms need not always address it to be falsifiable.

This brings us back to Naturegirl's original suggestion, which was a neural activity-based fishing expedition. We do these all the time, especially in bioinformatics, but not to generate models (and, therefore, nothing of interest to behavioral scientists, at least in the way you've proposed) It's like GWAS in that way. The point is that, since an AI is built to classify an input set into an output set (sort of) and her original post didn't include an output set, it's not the tool to use to pull out trends. We have better tools for doing that, PCA being one of them.

The original idea was to get a whole bunch of neural data together and learn from it. That's not inherently unscientific, but there are very specific ways in which it can be a useful part of the scientific process and AI can't really help with them. What we do with a giant pile of data, be it genomic or neural or proteomic or anything else, is fish through it for correlations. You can then go look through existing models to find ones your data may be able to disprove, but you don't go hare off and propose one ex nihilo just from fishing. At most, it's a source of very specific preliminary data. Useful, but not as a deliverable.

[wierd]

In both of my examples involving the ball, a cause is not asserted, only an outcome-- Ball moves in direction of club.


You are reading too much into my statement.


More specifically, rationalwiki states zetecticism is this:

"Zetetic" is an obscure English word coming from Greek through Latin. As an adjective it means "proceeding by inquiry; investigating", and, thus, when used as a noun - "inquirer". It has been used as a something-like-synonym of "skeptic" at least twice.

In the 19th century, the word was used by Flat Earth advocates: Samuel Rowbotham (under the pen name of "Parallax") wrote an anti-round-Earth pamphlet called Zetetic Astronomy and later founded Zetetic Societies in the UK and the USA and edited The Zetetic and Anti-Theorist: a monthly journal of practical cosmography. After his death, Lady Elizabeth Blount established a Universal Zetetic Society that was succeeded in the middle of the 20th century by Samuel Shenton's International Flat Earth Society. The word appears to still be popular among modern-day flat-earthers[1].

In the Flat Earth sense, the term refers to flipping the scientific method on its head and deriving one's observations from testing, with no regards to any hypothesis. Of course, if you did scientific inquiry this way, you'd end up with stating that a sphere is flat just because it looks flat to a relatively minuscule observer on its surface.

The original name of the Skeptical Inquirer magazine was The Zetetic. After Marcello Truzzi's falling out with CSICOP, he appropriated the term "zeteticism" for his brand of "more open minded" skepticism. He started another journal, the Zetetic Scholar.

Literally, question, THEN test.

EG--

"Does my ball go in the direction of my club when I strike it?"  made prior to any observation, or structured hypothesis of an outcome.

This is opposite to:

"I notice that the ball goes in the direction of my club. Is this ever false?"

which has the hypothesis about the outcome first, based on observations.  It's a test of the observed outcome.


The reason why the zetectic method is bogus, is because it is essentially a mutant form of begging the question, which presupposes that a question is correct, and then proceeding from it.  Instead, you start with an observation in science-- something outside which can be measured.  That thing exists regardless of the whim, whimsy, or desire of the experimenter, which is what makes it empirical. 

EG, the flat earther will go "Is the earth flat?"  then go "If the earth is flat, then this will be true."  (tests for thing, finds it true-- Concludes earth is flat.) 

This is opposition to real science-- "The earth appears to my eyes as flat. I will test if this observation is correct. To do so, I will devise a test that is sure to return positive if my observation is false."

 

In any event, the statement I made, (that you assert wrongly is zetecticism), is that the data collected (which is an observation made in absence of any causal theory applied-- it is raw observation) can be used to create a model.

The scientific method's first step is observation for a reason Trekkin.

    Make an observation.
    Ask a question.
    Form a hypothesis, or testable explanation.
    Make a prediction based on the hypothesis.
    Test the prediction.
    Iterate: use the results to make new hypotheses or predictions.

The statement that the data collected (which, lacking any original intent except to collect that information) can be used to make a model, is fundamentally correct.  It is the starting point for the scientific method.

[nogoodnames]

Unrelated random thought, what if we programmed an AI, and had it watch the brain activity of various humans throughout say, a college campus? To learn about the way humans think?

Biggest hat trick I can think of with that would be getting the whatever the brain activity data. If we have particularly portable devices capable of getting decent neural activity reads I don't think I've heard of it, and a quick google check suggests a: we don't, and b: the closest we have to it looks like stuff straight out of a sci-fi b-movie.

Well, there is this: https://choosemuse.com/what-it-measures/
Very crude and noise-prone, but I could see someone getting a research grant to make some undergrads walk around with these. 4-5 EEG channels probably won't give you much insight into "the way humans think" but it could provide a decent measure of how brain activity varies throughout the day. Might be useful for planning class schedules.

[dragdeler]

-

[Yoink]

The tips of one's fingers should be called the 'fingerheads'.

[McTraveller]

In another thread someone mentioned "alternative facts" and it made me think of a way to really make alternative fact opponents mad:  "6 plus 6 is twelve".  Alternative factual versions of that are "6 plus 6 is a dozen" or perhaps "6 plus 6 is C" (in hexadecimal) or perhaps even "6 plus 6 is 15" (in base-7).  See!  Infinite alternative facts!

(I feel a little dirty, to be honest.)

[dragdeler]

-

[Naturegirl1999]

youtube recommended this to me. The video is called Nuggets. It’s...interesting

Unrelated: I wonder why a tic tax toe tie is called a cat. Here is what google said

A tie in Tic-Tac-Toe is also called a scratch, which is short for cat's scratch. This comes from the idea of a game that is played perfectly will have no winner. It's like playing with a cat. When both players are at a stalemate, the cat usually signals its finish with a quick scratch.

This still doesn’t make sense to me. How is a tie like playing with a cat?