The Social Lives of Robots

Sunday, February 18, 2024
First Aired: 
Sunday, November 14, 2021

What Is It

Machines might surpass humans in terms of computational intelligence, but when it comes to social intelligence, they’re not very sophisticated. They have difficulty reading subtle cues—like body language, eye gaze, or facial expression—that we pick up on automatically. As robots integrate more and more into human life, how will they figure out the codes for appropriate behavior in different contexts? Can social intelligence be learned via an algorithm? And how do we design socially smart robots to be of special assistance to children, older adults, and people with disabilities? Josh and Ray read the room with Elaine Short from Tufts University, co-author of more than 20 papers on human-robot interaction, including "No fair!! An interaction with a cheating robot."

Part of our series The Human and the Machine.

Listening Notes

Can robots learn to read social cues? Is it possible to derive empathy from an algorithm? Ray questions the necessity of socially intelligent robots for jobs that humans already do well. Plus, she is skeptical about how numbers and computing data can create something as complex as social intelligence. Josh, however, argues that it is useful for robots to learn how to read social cues because they must navigate different environments and spaces.

 

The philosophers are joined by Elaine Short, Professor of Computer Science at Tufts University. Elaine’s work focuses on robots that help and learn from people, as well what happens when robots exit the lab and enter the world. Ray asks if robots are truly learning social intelligence or if they’re simply simulating humans, but Elaine considers the distinction to be unimportant in her field. Josh asks about the success of companionship robots, which leads Elaine to describe the success of animal and zoomorphic robots. She believes that humanoid companionship robots will still take time to develop, especially since a large problem in social robotics lies in managing human expectations.

In the last segment of the show, Josh, Ray, and Elaine consider the tension between popular science and sci-fi representations of robots versus how they actually operate. They look at various weaknesses of socially assistive robots, such as their potential to make mistakes, accidental emotional harms, accessibility, and high costs. Elaine emphasizes the importance of increasing diversity in robotics and computing, and she explains how assistive robots can aid in disability rights and empowering people with disabilities.

  • Roving Philosophical Report (Seek to 4:27) → Holly J. McDede discusses how a robotic bee and a robot designed to help kids with autism spectrum disorder are impacting the social lives of their respective communities.
  • Sixty-Second Philosopher (Seek to 49:01) → Ian Shoales examines the diversity and tropes of robots in pop culture.

Transcript

Transcript

Josh Landy  
Could a robot ever really understand you?

Ray Briggs  
Could it at least help your kid s play well with others?

Josh Landy  
Would you trust a robot to take care of your child?

Comments (13)


Harold G. Neuman's picture

Harold G. Neuman

Saturday, October 2, 2021 -- 12:19 PM

They need not have social

They need not have social lives. Because they do not think...no matter what anyone tries to tell you. AI is contrivance. Turing would have told you that. Actually, he did. But no one was paying attention.

I've read and agree to abide by the Community Guidelines
Tim Smith's picture

Tim Smith

Thursday, October 7, 2021 -- 7:11 PM

Turing said and did many

Turing said and did many things and was ignored for various reasons, but to characterize his view of AI as a contrivance and that no one noticed is like saying Barack Obama was a racist and Donald Trump loved his country and we all missed it. All these statements are valid, but all are equally untrue.

This is a cool little sub-thread, though. Alan Turing won WWII, was one of the greatest philosophers, mathematicians, and human beings of the last century… PT should do a show on him.

I've read and agree to abide by the Community Guidelines
Harold G. Neuman's picture

Harold G. Neuman

Sunday, October 3, 2021 -- 7:34 AM

I think professionals get so

I think professionals get so wrapped in their disciplines, they are not able (willing?) To separate reality: 'how things probably are' from fantasy: ' how they might possibly be'. Sure, I tend to think outside the box and/or jump out of the system on a few things myself. It is good mental gymnastics;helps keep me sharper in advanced years. But, when we talk of notions such as computer social skills---or the lack thereof, we are in the realm of What Does it Matter?, not how things probably are. This is more Asimov than is either practical or pragmatic. Is it useful to speculate on the improbable? I don't think so. Artificial Intelligence is making differences-this much is inarguable. Mental gymnastics are good therapy. And, idle chat IS a social skill, banal as it can often be. Computers are marvelous tools, SKA, algorithims. I would, however, rather play chess with another human being.

I've read and agree to abide by the Community Guidelines
Tim Smith's picture

Tim Smith

Thursday, October 7, 2021 -- 9:13 PM

Exoskeleton tech can get

Exoskeleton tech can get people with paraplegia mobile. Assistive robots have infinite patience for autistic kids to learn from and play with even. Eldercare robots give failing parents peace of mind. Medical rescue robots provide people with epilepsy dignity and timely care. There are way more applications than this for robots. These roles call for social skills and sensitivity to perform well. Philosophers must consider the limits and possibilities.

Advanced neuromorphic chips are in the pipeline to add neural networking units to process the cues that could detect a seizure, log depression, and find avenues to engage a young autistic mind. These and many other tasks can be done and done right. It only takes a few fails for people to lose their trust in tech to solve and address these high-need use cases.

I agree Robots don’t need to be social like humans, but they need to respond to human emotion and traits (some of which humans might not even perceive.) The most dangerous thing a robot could do would be to imitate humans, perhaps. Sometimes you want a robot to act and feel like a human; in general, robots should communicate socially with humans that they are not themselves humans. If that seems odd, it won’t very soon. Robots are becoming more and more human. That, I agree, is a problem.

I've read and agree to abide by the Community Guidelines
Harold G. Neuman's picture

Harold G. Neuman

Wednesday, October 20, 2021 -- 5:02 PM

Tim:

Tim:
I do not often look to film or popular literature for philosophy.. But, I was moved by I Robot. The hi-tech Audi was neat, too.

I've read and agree to abide by the Community Guidelines
Tim Smith's picture

Tim Smith

Wednesday, October 20, 2021 -- 5:55 PM

Harold:

Harold:
"The Imitation Game" is worthy wrt Turing.

I've read and agree to abide by the Community Guidelines
tinkwelborn@mac.com's picture

tinkwelborn@mac.com

Tuesday, November 2, 2021 -- 11:06 AM

Nice banter, here.

Nice banter, here.

I've read and agree to abide by the Community Guidelines
Daniel's picture

Daniel

Saturday, November 6, 2021 -- 6:14 PM

And informative! Not clear

And informative! Not clear to me why more scholars are not making use of this excellent resource, which I describe as a public utility of intellectual exchange. At any rate, the notion of socialized robotics seems closely related to that of artificial intelligence, a term conspicuous for its ambiguity. If intelligence consists in the capacity for discrimination among relevant alternatives, e.g. in the case of a motion sensor signaling a person's body entering a structure rather than a flurry of leaves blown in by the wind, there's no detectable difference between artificial and natural intelligence. If, however, as forum participant Smith suggests in the first paragraph of his October 7 post above, artificial intelligence picks up where that of normal intelligence leaves off, then it's a continuation of the latter, and therefore distinguished only by degree. But the concern many have over how such technology is used in our society I think is very important. That is that artificial intelligence constitutes a consumer product of natural stupidity. Take self-driving cars. A product of natural intelligence in relation to existing conditions in the U.S. would be to reduce the economic externality of traffic congestion and damage to the atmosphere by hydrocarbon combustion by means of a rapid transition at the national level from private to public transportation. To represent artificially operated private transportation as a kind of intelligence would only then be plausible insofar as natural intelligence plays no role in its voluntary deployment.

I've read and agree to abide by the Community Guidelines
Devon's picture

Devon

Wednesday, November 17, 2021 -- 4:35 PM

Fanya in Aptos, CA sent us

Fanya in Aptos, CA sent us this question:

"Would you please ask the guest whether she's thought about using what we know about mirror neurons in her research to get a robot to model human beings and respond to humans?"

Unfortunately we received it just a bit too late to include in the program, but our guest Elaine Short sent this response:

"HRI researchers are often inspired by human cognition, including mirror neurons! In our work, that might look something like modeling how the robot would do a task while watching a person do that task. So a robot trying to figure out how to help someone put groceries away might make a plan for how to put away the oranges, which the robot is holding, AND a plan for how to put away the pasta, which the person is holding. It can then use the plan it made "for the person" to either detect when something surprising has happened ("The person put the box in the freezer! Maybe it was a box of mochi, not a box of pasta") or to make better decisions about its next actions ("The person is going to put the pasta in the cabinet, which means opening the cabinet by the fridge, so I should wait before I try to open the fridge")."

I've read and agree to abide by the Community Guidelines
Tim Smith's picture

Tim Smith

Thursday, November 18, 2021 -- 9:50 PM

Mirror neurons were

Mirror neurons were discovered in macaques. The only problem is … macaques don’t imitate. The fundamental papers on this were straightforward and unassuming. Later, mirror neurons were proposed to explain language, theory of mind, and imitation based loosely on a debunked theory of motor learning in speech.

An excellent book that explains this is ‘The Myth of Mirror Neurons' by Gregory Hickok. I read it a few years ago and can’t remember all the points. Still, he pretty much squelches any hope of using the mirror network to do anything other than help explain rudimentary motor/action function, which may be the use case here.

There are lessons to learn from human consciousness. These are not lessons for robots or artificial intelligence, for the most part.

As brought out in the show, the more common issue is that humans assume a robot is social, that it can learn and approach human expectations. As Elaine said, that would require good old-fashioned AI to accomplish at this point, and maybe all points. There’s too much goodness in robots to waste time living up to false and unrealistic expectations. Currently, we should focus on safety, ease of use, and easily automated assistive algorithms for high-need use cases.

I've read and agree to abide by the Community Guidelines
Harold G. Neuman's picture

Harold G. Neuman

Sunday, January 16, 2022 -- 1:53 PM

Whoa, now. This choice of

Whoa, now. This choice of subject matter implies that not only could robots be people, they are. A bit premature, don't you think? I am not there yet. I might countenance a notion of life extension through improved circuitry if only I could see some percentage in it. But I am not so self-indulgent (or self-assured) to believe extension of my life would make a difference that made a difference. That would amount to loathsome egomania---which we have seen more than enough of. I like to speculate about an array of topics. I even like dancing robots. Humans are social animals, when they choose. So are cats. Society; social; & socialization are words and concepts that fit certain life forms more precisely than others. Unless the definitions of the words and concepts have changed, they do not fit machines---even sophisticated ones. (Please. Make no mistake. AI and robots are machinery.) Those who elect to use such words and concepts in discussions around AI, robots and so on, need both a faster horse and better cart to put before it.

If this is absurdity, I'll take it over contextual reality. The next big thing is notoriously illusive.

I've read and agree to abide by the Community Guidelines
cgptonline's picture

cgptonline

Sunday, February 25, 2024 -- 7:56 PM

In the age of advancing

In the age of advancing technology, the intersection of robotics and social intelligence raises intriguing questions about human-robot interaction. While machines excel in computational tasks, deciphering social cues remains a challenge. ChatGPT Online could contribute to exploring algorithms aimed at enhancing robots' ability to navigate complex social landscapes, bridging the gap between artificial and human understanding.

I've read and agree to abide by the Community Guidelines
Vladimirbalistreri's picture

Vladimirbalistreri

Saturday, April 27, 2024 -- 2:50 AM

fnf mods is a fascinating

fnf mods is a fascinating website devoted to Friday Night Funkin' game series. In this series of games, players are challenged with loads of wonderful music, rhythms and sounds that keep players engaged without getting bored.

I've read and agree to abide by the Community Guidelines