ALGORITHMICALLY Detecting Autism From Forum Behavior

Post Reply New Topic

Let's say there's two kinds of people who are on an autistic forum for ease of analysis.

There are those with Autism and those who don't have autism.

What are the forum behavioral characteristics of those with Autism? What are the forum behavioral characteristics of those who are neurotypical?

What are some universal communication trends and how can we apply those to algorithmically determining whether someone is likely autistic or neuorytpical?

Addressing the last question, there's a well known phrase in the sociological world and is observable in everyday life that goes "Birds of a Feather Flock Together". What this means is that people with similar communication patterns, similar interests and so on tend to flock together. Assuming the existence of interests, concerns and ways of thinking that are more common among NTs than aspies or vice versa, then NTs would tend to be more common in threads in which NTs and Near-NTs are participating and vice versa for Aspies and fellow aspies and near-aspies (Like BAP). Private message communication would also show this kind of similars-attract trend.

What are some other trends?

Those with a more well-developed ToM and would have a greater ability to point out things of interest to fellow forum users, and assuming those who are fundamentally NT would tend to have a better ToM, then NTs would be more likely to attract more responses and essentially have more popular posts and/or threads.

So, in algorithmically analyzing the presence of autism among forum users, two useful metrics to measure would be popularity and seeing who's responding to who, and determining the overall composition of ones virtual "social circle". Those who are NT would tend to have more NT social circles and those who are Aspie would tend to have more aspie social circles.

How can we infer one's social circle?

Let's say that we define a friend or "associate" as someone with three or more mutual responses with another user in a sufficient number of threads. In other words, if two users have enough "conversations" in enough threads, then there's likely enough similarity for them to be considered "like minds" or "Near like minds" in some way.

So, now I need a copy of the website's forum database to create a program that analyzes the social behavior of the users and determining who's more likely neurotypical and autistic. I could create a web spider to download a large part of the forum and I could build a parser in order to generate my own database for analysis, but it'd be far simpler if there's someone who could send me a copy of the forum database, even though I know how unrealistic that is.

i thankyou havben watch ingg a new seris off Numbers?

Yes, you create a mthematical likelihood for nearly anything...

My algorithms for your behaviour are easily achieved due to your large output, including mitigating dishonesty and farce of course

An algorithm is a set of instructions, sometimes called a procedure or a function, that is used to perform a certain task. This can be a simple process, such as adding two numbers together, or a complex function, such as adding effects to an image. For example, in order to sharpen a digital photo, the algorithm would need to process each pixel in the image and determine which ones to change and how much to change them in order to make the image look sharper.

Most computer programmers spend a large percentage of their time creating algorithms. (The rest of their time is spent debugging the algorithms that don't work properly.) The goal is to create efficient algorithms that do not waste more computer resources (such as RAM and CPU time) than necessary. This can be difficult, because an algorithm that performs well on one set of data may perform poorly on other data.

As you might guess, poorly written algorithms can cause programs to run slowly and even crash. Therefore, software updates are often introduced, touting "improved stability and performance." While this sounds impressive, it also means that the algorithms in the previous versions of the software were not written as well as the could have been.

The less responses to your thread starts, the less friends you make on WP, the worse your grammar and written flow, the longer your boring paragraphs and an excess of wordage, the more aspergian you are

Someone like Moog, who barely constructs a whole sentence, and has many ardent followers, may indicate a complete lack of autism

swbluto may in fact be an intellect whore, or her attention needs are higher than most, along with a soft spot for meaness....she also trolls the newbies in the general forum, as the regulars have wised up to her aloof ambiguity, hence my prognosis of ADHD

That will be $540 thank you

Dont forget your meth based meds and please call the next patient

Dude, I can't believe you've been a member here for 8 months and still don't know if you're an aspie. Do you think you should maybe get tested, or admit you don't have it? I come here for information as I think that's what the site is mainly about. But all these posts about "do I have it", "don't I have it", "can I analyze everything everyone has ever said", you've got monkeys balls.

Dude - maybe you're a quantum aspie. You behave like an aspie as long as you don't try to measure it. But as soon as you attempt to quantify it, you collapse into an NT.

I don't feel terribly at home on this forum actually, I get information on here and try to give advice where applicable, but that's about all. If there's people I identify with on this forum, it's the over 30 people, but that's all. I'm NVLD, and Aspergers is only a comorbid or possibility for me. NVLD is kind of rare, though, supposedly only like 1/1000 people are diagnosed with it, and it's more like 1/100 for Aspergers.

That said, I feel much more at home on car forums, with posts like this all the time:

What you are trying to do reduces to a Turing Machine that halts. This requires precision in definitions and process that do not exist within the current body of autism knowledge. You can create a psuedo-algorithm, but it will be fraught with exceptional cases that undermine its validity.

And it still won't tell you if you are on the spectrum because you EXCEL at finding exceptions.

Interesting assertions. Do you have any hard data to back them up, or are you (like many sociologists) speaking through a non-standard orifice?