An AI based solution

[Sam39]

What do you think about an AI based solution that helps to communicate with autistic people ?

 

[LogicalMind]

I think people really need to have real people to speak to... AI is a great idea, but at the end of the day, it won't be complete enough to have complete human mimicry...will fall short... Might be a good thing for low func autos to learn social ettiquette and responses , but won't create bonds or emotional sustainability for anything more

 

[The Midge]

What do you think about an AI based solution that helps to communicate with autistic people ?

Like this:

?

 

[Sam39]

The device actually uses a camera and some sort of sensors in order to analyse the situation and its impact on the autistic person and then send a message and an advice to his parent on how he should do and then the device get a feedback and learn how to the parent should interact.

 

[OlLiE]

i guess the first distinction would be if they are high functioning or not

personally, for adults: getting to 'understand' anyone is always difficult (ASD/ NT),
everyone has their history, their hangups, their sensitivities which are harder to detect than their strengths without actually investing in these people to try to get to know them and to understand why they are the way they are and why they react the way they do in words and facial expressions

i'm a great fan of tech, gadgets and ai, for me however it stops when it tryies to replace the human effort of one person investing in another to get to know them, i find it lazy

personally, for a child, put in the time, get to know it, interact with it, grow with it, look for the signals, make mistakes and learn from it, with an ASD child it may take longer, so what..

personally, i think the effort is part of what makes a friendship or a parent child relationship, anyone wanting to use ai shouldn't have bothered having trying to have friends or children
, just talk to siri all day long for company...

i could see a use for it when it comes to interactions with random people that you never see again, some kind of haptic reaction to your watch when the system detects that the other person is moving away from neutral or happy

 

[Streetwise]

welcome shalom I'm not Jewish

What do you think about an AI based solution that helps to communicate with autistic people ?

 

[LogicalMind]

Problem with autistic people...is that you cant measure emotions through imagery xD

 

[OlLiE]

Problem with autistic people...is that you cant measure emotions through imagery xD

i've heard that, but if you can tell the difference between black and white, you can recognise the difference between a smile and frown and through association with experiences (and TV) learn that frown is bad, smile is good

i don't need to emotionally identify to recognise signs

 

[OlLiE]

at the end of the day on a face you have a few indicators (forehead, brow, eyes, nose, lips, loudness of voice) and can work off an algorithm

purely empirically: if this, then that

when i did the screening i couldn't recognise the emotions because i couldn't grasp the context, by getting older i've created a library of 'contexts' which can potentially lead to certain emotions, which gives me context

you can even learn in what situations a certain expression of emotion can have the opposite meaning of what it generally is, pe crying out of happiness, pushing someone away out of love etc

i was diagnosed through a three day screening process, the result was that i was clearly on the spectrum, that doesn't mean that i can't learn through observation no?

 

[LogicalMind]

@OlLiE I frown when I am excited or happy, and smile when I feel angry or ludicrously erratic... My voice is aggressive when I am calm and quiet inside, and soft when I am ... not good.

Simply telling you that there are statistical outliers in terms of indicators...and using facial recognition for asperger's emotion identification is less reliable than asking EA to bring out a good strategy game, or taking aspirin when you have low blood pressure xD

Your premise of using visual identification is good, but with us autistic people it is near impossible, or at least for the upper 20% who dont exhibit similiarities (alexathymia)

I am someone who has coded minor soft AI stuff with voice synth, words in 2012...and I can tell you... Emotional facial recognition aint gonna work on alexathymic, or social anxiety groups

 

[OlLiE]

@OlLiE I frown when I am excited or happy, and smile when I feel angry or ludicrously erratic... My voice is aggressive when I am calm and quiet inside, and soft when I am ... not good.

Simply telling you that there are statistical outliers in terms of indicators...and using facial recognition for asperger's emotion identification is less reliable than asking EA to bring out a good strategy game, or taking aspirin when you have low blood pressure xD

Your premise of using visual identification is good, but with us autistic people it is near impossible, or at least for the upper 20% who dont exhibit similiarities (alexathymia)

through observation i've learned to detect emotions, and i've taught myself the appropriate reactions and facial expressions, i can fake emotions in social contexts, but i have to stay in bed most of the day after, i only have a few people that i have studied enough to feel comfortable interacting with and making mistakes with, most of the people i know where surprised when i told them my diagnosis

 

[LogicalMind]

through observation i've learned to detect emotions, and i've taught myself the appropriate reactions and facial expressions, i can fake emotions in social contexts, but i have to stay in bed most of the day after, i only have a few people that i have studied enough to feel comfortable interacting with and making mistakes with, most of the people i know where surprised when i told them my diagnosis

I agree with you...but I mean that the computer cant read our facial expressions... I dont exhibit NT facial expressions for any of my emotions, I really look the opposite of how I feel, and the computer wont know that. Maybe I miscommunicated my premise ineffectively lol...but I know Alexathymic people can learn to read emotions, but a machine cant read or be programmed to read Alexathymic people's emotions...that is the coveat/crux

 

[OlLiE]

@OlLiE I frown when I am excited or happy, and smile when I feel angry or ludicrously erratic... My voice is aggressive when I am calm and quiet inside, and soft when I am ... not good.



actually, when i think about it, i guess an ai would use a similar algorithmic process as i've described,

at the end of the day an AI is basically a statistical model, i.e. find a representative sample, determine the statically relevant drivers, add scores that differentiate an 'intensity', create a formula with an overall score output, correlate the quantitative output to a qualitative conclusion, test against a random sample, hey ho presto

 

[OlLiE]

I agree with you...but I mean that the computer cant read our facial expressions... I dont exhibit NT facial expressions for any of my emotions, I really look the opposite of how I feel, and the computer wont know that. Maybe I miscommunicated my premise ineffectively lol...but I know Alexathymic people can learn to read emotions, but a machine cant read or be programmed to read Alexathymic people's emotions...that is the coveat/crux

i agree it can't 'read them', but based on a large sample, if you could define facial variables and find a way to quantify their degrees of variance, i.e. 'quantify' facial indicators, then define a baseline neutral face, if you can then correlate the deviance data to an emotion then an ai could , with a certain statistical degree of accuracy, stick an emotional tag to the data it has received, after that you would have to statistically calibrate the model to an individual face, no?

i'm not an IT guy by the way, but was active in creating predictive credit risk models

 

[OlLiE]

sorry by the way, i was so interested by the topic, that i switched things around, i previously posted something in another thread, 'things to invent' where i suggested a tool to help us understand NT's, sorry, my bad

the other way round would be harder because of a lack of data regarding the one aspie you want to understand/model

apologies

 

[LogicalMind]

i agree it can't 'read them', but based on a large sample, if you could define facial variables and find a way to quantify their degrees of variance, i.e. 'quantify' facial indicators, then define a baseline neutral face, if you can then correlate the deviance data to an emotion then an ai could , with a certain statistical degree of accuracy, stick an emotional tag to the data it has received, after that you would have to statistically calibrate the model to an individual face, no?

i'm not an IT guy by the way, but was active in creating predictive credit risk models

You would do a brilliant job from the heuristics point of view... it could capture 99.998% ...but the problem is the remaining fraction are outliers that dont exhibit any statistical convention... trying to manual go through a sample of lets say ten thousand disparant asperger videos, and somehow trying to quantify it and capture it into an xml file or database, would make me #### my pants xD lol... I am a masochist, so I would enjoy it, but it is ludicrously infeasible... Like you say though, if we establish a smaller target audience (low func autism) we may be able to feasibly do it, since they dont have that many outliers...

In theory you are 100% right, anything is possible... in practice, it would take you ten years to cater for all asperger emotions before the program is effective enough (sample data, heuristical maps, performance issues from processing huge sources of data which isnt indexed since they are outliers) feel sorry for the IT j@ackass who volunteers for that project lol...he is gonna cry

Although, this all depends on what the scope of the project is? I was talking about conversational AI that would pass a touring test

 

[OlLiE]

as an aspie, I can live with a 10 % margin of error

nice to exchange ideas with you

 

[LogicalMind]

You are very clever OlLiE... and I failed maths in high school lol... I enjoyed this discussion, it warmed the cockles of my heart, and made burdened my groin and bladder...need a bathroom break now xD

 

[LogicalMind]

sorry by the way, i was so interested by the topic, that i switched things around, i previously posted something in another thread, 'things to invent' where i suggested a tool to help us understand NT's, sorry, my bad

the other way round would be harder because of a lack of data regarding the one aspie you want to understand/model

apologies

Understanding NT's via some image processing is definitely feasible... That could easily be done, so you are on the money with that one man...you got the gold hens there

 

[OlLiE]

ps re low function autistic children/people - maybe physical detectors would be more useful - it would be great if you could make an 'mri cap' that can interpret "emotion" data in the brain, that way you could bypass the need for facial detection and ai altogether

an ai will always be dependent on a statistical model, no data, no ai