What is the role of psychological knowledge in designing AI solutions

Written by Axel Rosenberg

Many say that Artificial Intelligence reaching singularity is just around the corner. Around the corner is something that some say is 5 years and some say 100, but quite possibly in our lifetimes. Artificial Intelligence reaching singularity, or becoming an Artificial Super Intelligence (ASI), is expected to have extreme changes on society and its construction, while also being speculated to be the next step in evolution and making humans extinct.
I think Lieutenant Commander Data, an android from Star Trek, is an interesting take on Artificial Intelligence that I find quite soothing. One could argue that Data and his intelligence has achieved singularity, but in the series, he struggles with human (and other races) emotions. His greatest ambition is to be human himself, even though he is both stronger, faster both cognitively and physically, smarter, has seemingly infinite memory capacity with the ability to recall anything, and is immortal.
Data acknowledges that he can not feel emotion, but he still tries to imitate what he sees other officers doing, often failing miserably. While many could argue that he is a superb being and has surpassed humanity, his struggle to achieve humanity is interesting to watch.
He does, however, have psychological knowledge and understands what emotions are. He thinks of everything objectively, but still often fail to make sense of relations between emotional beings. His objectivity does however provide great advantages, and not having emotions gives him clarity in difficult situations.
In the setting of a military command, an objective android with AI is what the command needs and requires. In some cases, however, an emotionless objective AI can be disastrous. For our project for this course, we designed a chatbot meant for individuals suffering from mental heath issues. Severe cases can have extreme situations which can escalate very quickly to something disastrous, so an AI behind the chatbot must understand and react to emotions and psychological aspects correctly depending on the situation. It also must know when its skills are not enough, and when a trained professional is needed (or at least for now, since no chatbot has ASI capabilities.
In any case, the designer has to have a basic knowledge about psychology. Preferably there is a trained psychologist in the team that is designing and advanced AI, but context is key. Some AI solutions requires psychological skills while some don’t. I am however sure that humanity will some day create and AI that evolves itself in such a way that human development of the AI is no longer needed.

How AI Hub is bringing us closer to robot friends

Written by Miikael Lehtimäki and Ivan Dubrovin

AI Hub Tampere is a new research center hosted by Tampere University and will bring together people from all aspect of AI research and bring ai research closer to robotics and other uses. The teaching goals of the center are to train people to develop AI, engineers to implement AI and managers to outsource AI. The primary disciplines brought forward are visual imaging, machines learning and audio & signal processing and with experts of these areas working together they can all take strides together. 

With new advancements in visual imaging coming forth AI becomes more capable of visually recognizing things, being able of recognizing people, specific pieces of clothing and other equipment along with reading emotions from faces. This comes in use for helping social robots read situations and feelings of their subjects, the ability to read items, people and faces together can help recognize suicide or other attempts at self-harm along with being able to play with toy or devices together with people. 

With individualized solutions for tasks and subject using the new algorithms and solutions, including ones which enable learning from example and two-handed manipulation of objects, robots and devices can learn unique ways to accomplish tasks for unique situations, like taking time and its passing into account knowing what is recent and what that means. Solution that can come from this are throwing balls with two different children, the robot can start with a general solution with the height and reach of the child and specialize its throws for each to be catchable and if the child threw balls last a long time ago, start by going easy on them, learning what improves the mood of their owner, this combined with data sharing between platforms can allow robots to try new or different things to solve new problems, a robot that sees that you are in a bad mood tries cartwheels, bringing you a ball, singing or playing a song or sound for example. 

Advancements in audio and signal processing allows for the detection and categorization of individual sounds, from speech, birds, cars and being able to isolate these sounds out from the others. With these improvements robots can use simple audio receptors to become far more aware of their surroundings, recognize people more accurately and listen to multiple people at the same time, improving social robots used in groups as they will get less confused from crowds, able to recognize authorized commands from a specific individual even in loud situations and better able to read situation context as in a case of their owner playing sad or happy music or their tone of voice.

Alone these directions can already bring great strides to social robotics in general but together they bring us ever closer to the robot friend, an ai companion that can learn to know you, knows what cheers you up, suggest things to do, complains about things that annoy it and can tell you to get them maintenance. Less a creepy android servant and more a companion device that can journey through life together with you in this ever-expanding digital world.

What about technology for elderly?

Written by Joonas

What do we all have in common? We are all aging… Technology isn’t only used by people in their 20s or 30s. The population (in Finland) is getting older and older [1]. We shouldn’t exclude the elderly from the use of technology. After all, most of us are considered as elderly people at some point.

When we think about the elderly, the first thing that comes to mind is the decrease of the senses of vision and hearing. Also, the decrease of the working memory capacity and motor skills. All of these put the elderly in disadvantage against the younger generation. Loss in near vision can be corrected with glasses, but reduced field of vision, decline in contrast and colour sensitivity and motor skills cannot be corrected as easily. Fortunately, this can be made easier for the user with proper design guidelines. Technology should be accessible to everyone, no matter the age or abilities of the user. This can be achieved by devices designed specifically for elderly, or by designing the content with accessibility[2] in mind.

The elderly require usually assistance (and persuasion) in taking technology into use. More often than I can remember, I have said that “You should try to do it first by yourself. You cannot learn to do it, if you don’t try” while simultaneously trying to teach how to do the task. Usually they haven’t even tried to do it by themselves before asking me to do it for them. Often, they succeed with the task with minimum help from me. Maybe they just wanted someone to oversee the whole situation, so they would feel comfortable with the technology.

If the technology product is branded for “the elderly” or as “assisted living aid”, it might make it hard for the user (in this case, an elderly person) to accept using it. They don’t generally want to be seen as someone who can’t take care of themselves. Even if the device is technologically advanced and full of features, it might still be considered as an elderly product and left to be. I think this is nowadays the biggest challenge when designing technology for elderly. After a while, the situation could be entirely different because the next generation has already used all kinds of devices for a long time.

References:

TIE-40106 Psychology of Pervasive Computing Lecture 12 – 19.11.2018

[1] https://www.tilastokeskus.fi/tup/suoluk/suoluk_vaesto_en.html

[2] https://www.w3.org/TR/WCAG21/

Social robot, could it adapt to every situation?

Written by Antero Mäkinen  

Could there be a social robot, which fits for all humans? All-in-one robot? I don’t think so, or at least it should have some transformer’s skills to adapt to every situation. Human’s gender, age and senses matter very much when concerning how social robot fits for every individual human. Some people don’t see, some people are deaf, some are young, and some are old and so on. I watched a couple of videos from YouTube to think about how user group matters when designing robots. A video called “The next frontier in robotics: social, collaborative robots | Andrea Thomaz | TEDxPeachtree” (https://www.youtube.com/watch?v=O1ZhWv84eWE) explained it quite well with great examples.

In this video, Roboticist Andrea Thomaz had her robot buddy with her and she showed couple of examples with the robot. Even though the video is almost three years old, I think it’s still valid. I was really surprised when her robot started to talk, and it sounded like a little kid even though in my mind it didn’t look a kid, more like a monkey, so maybe it should have sounded a bit monkey-ish.

So, the first thing I would consider when thinking about designing a robot is that how it should visually look and what kind of voice should it have to match the visuals. For example, maybe for kids or elder people it wouldn’t be so weird if the robot talked like a little kid, but for middle-aged people it probably wouldn’t be so good choice if they wanted to take the robot seriously. Of course, there are always exceptions and good thing is that robots can have multiple voices in the same package.

Second thing that needs to be considered when designing to different age groups are gestures. Gestures are quite common in human-human interaction and in the video example, the robot acted very inhuman when she did actions without showing her eye and/or hand and body movement. This would definitely scare youngest and oldest users who would interact with the robot, because it would act so differently compared to humans and thus it wouldn’t be so easily approachable. In industrial robots, which actually doesn’t even try to look like humans, these things aren’t so important, but with social robots they truly are.

Last thing I would like to point out is the need of human-kind of interaction. It includes everything: visuals, gestures, sounds and even behavior. I think social robots should show empathy for all target users, but especially for kids and elders or those with special needs. First level of being empathic would be just to say correct things at correct times, but I think we’re still quite far from empathically skilled robots. Besides words, robots could also show emotions via gestures, for example by showing their support to grief they could give hugs or simply touch user’s shoulder, like we humans do.

To summarize, there are quite many things that need to be considered when designing social robots. Creating fits for all social robot that would have perfect visual, audio and behavior experience to every user will be hard, but I’m looking forward what the future brings on this area. Will social robots adapt perfectly to every user and how would they adapt in a group situation with multiple humans and robots?