The Colorado School of Mines InteRactive RObotics Resarch (MIRROR) Lab performs cutting edge HRI research. Visit https://xitter.cf/t.co/78uqzzrrbn to learn more.
Golden, CO
Joined August 2017
- Tweets457
- Following88
- Followers424
- Likes257
The MIRRORLab will no longer be posting on X. The Neo-nazi and authoritarian sentiments it has come to represent and support are in direct opposition to our lab values.
Please follow our lab members on Bluesky and Linkedin.
You can get a sneak peek at Andres' work on this topic in his 2023 HRI LBR dl.acm.org/doi/abs/10.1145/3…
Keep your eyes peeled for more work from Andres in the future!
Thank you again to Andres for helping us close out our speaker series this year! /end
In the QA, we discussed key differences in human vs robot use of deception, the impacts of power and social status on perceptions of deception... and the brilliance of The Brothers Karamazov :) 6/
Andres also found interesting differences between conditions in terms of whether attributed deception-related blame to robots alone, versus to other humans viewed as being "in" on the deception. 5/
Andres found that robots' use of external state deception (white lies) was generally approved of, while others types were not. This makes sense since robot deception was most frequently justified by participants when it was viewed as being use to prevent harm. 4/
There's been some recent theorizing (from @JohnDanaher) about different types of deceptions robots can engage in -- and Andres work seeks to provide the first experimental evidence for different perceptions of robots on the basis of those types of deception! 3/
As a member of @phillips_beth's ALPHAS Lab, Andres is interested in how robots can give justifications (rather than mere explanations) for their actions. And in this work, Andres explores how people view robots' use of deception in the course of those justifications. 2/
Thank you to our final speaker of the 2024 summer, Andres Rosero of @GeorgeMasonU! Andres talked with us about justification, moral dilemmas, and robot deception. 🧵
And overall, you can learn more about Kantwon's work at Kantwon.com
Thank you Kantwon for speaking with us today!
/end
We particularly recommend Kantwon's HRI 2024 paper
dl.acm.org/doi/abs/10.1145/3…
But you should also check out the work he's been doing with the Rebellion and Disobedience workshop series! 8/
In the QA, we discussed a wide range of questions, including the key question we've all been asking... should robots lie to kids about Santa? And if so, why, and how? 7/
How can we scale educational LLMs to large classes? Can lowering confidence in CS be good because it helps reveal learning opportunities?
6/
Kantwon also shared insights from his recent research, where he's asking...
Can robot lies that are prosocial have equal benefit when delivered by robots?
Can playing dumb be effective when robots promote learning by teaching?
5/
3. That efforts to repair trust through apology can lead to more norm violation if they reveal differences in value alignment! 4/
In Kantwon's talk, he talked about a number of compelling research findings:
1. That lies from an AI system not seen as negatively as lies from humans!
2. That deception by robots significantly increases expectations of *malicious behavior* by humans!
... 3/
Kantwon's research asks a Big Question: Should robots lie?
As Kantwon points out, it's inevitable that robots will deceive. Understanding the nature of that deception and how it is perceived is critical for HRI. 2/🧵
We especially recommend Bengisu's HRI 2024 paper "Toward Family-Robot Interactions: A Family-Centered Framework in HRI" dl.acm.org/doi/abs/10.1145/3…, but you can learn all about Bengisu's body of work at bengisucagiltay.com/
Thank you again Bengisu for talking with us today! /end
And after her talk, we had some great conversations about Bengisu's work -- it was great to see how inspired our high school summer interns were about possible use cases for social robotics! 6/