Artificial emotional intelligence - Giles Colborne
71 likes20,004 views
The document discusses artificial emotional intelligence and designing emotionally intelligent interactions. It summarizes approaches like building rapport through active listening, empathy, and understanding users' motivations and personalities. While advanced emotional detection may be far off, simple techniques like tuning interactions to individual differences can improve experiences now. The document advocates designing flexible, responsive systems that adapt interactions based on dynamic understanding of users.
Artificial emotional intelligence - Giles Colborne
- 1. Artificial emotional intelligence Giles Colborne http://www.flickr.com/photos/thomashawk/54164255/ 1
- 2. I hope I’m not I’m a bad person. But sometimes I think bad thoughts. I want to punch that website in the face 2
- 3. This guy acts out those bad thoughts. He’s a bit scary. http://www.youtube.com/watch?v=HtTUsOKjWyQ 3
- 4. Something about computers brings out the worst in us. We shout at our laptops and throw our remote controls on the ground. 4
- 5. In response, some interaction designers act like Nurse Ratched in One Flew Over the Cuckoo’s Nest. They see emotion as the enemy. But I think that misunderstands the importance of emotions. 5
- 6. AI pioneer Marvin Minsky sees emotions as ‘ways to think’. Emotions tune the machinery of the brain. Even a ‘negative’ emotion like anger is useful. It gives us the energy to address a threat and sends out social signals that warn others we’d better get our way. 6
- 7. People without emotions are less effective thinkers. Antonio Damasio describes one patient, Elliot, with above average IQ but frontal lobe damage that means his emotions are impaired making him a kind of Mr Spock. When faced with a ‘rational’ task like scheduling an appointment, he endlessly weighs and compares choices. He doesn’t feel boredom, frustration or embarrassment. He takes for ever. Emotions are vital to decision making. 7
- 8. Why do we like to watch horror movies? Because they help us extend our emotional range and understand our emotions. Emotions are important to people. They’re linked to our strongest memories and the most meaningful events in our lives. What if we accept there So we’re right to reject will be ups and downs in the Nurse Ratched view the relationship between that we should eliminate humans and computers? emotion. What if we try to give our designs the emotional intelligence to ride those waves? 8
- 9. Marvin Minsky wants to create artificial emotional intelligence by building an artificial brain. Awesome! But that’s a few decades away. Meanwhile, I have a website to design. So are there ways I can cheat? 9
- 10. Back to this guy. The word that describes him is ‘postal’. So maybe we can learn how to deal with this situation by talking to some experts. 10
- 11. These folks train negotiators to deal with crisis situations. Maybe FBI they can help us. 11
- 12. Behavioural Change Stairway Model They use this model to get through hostage negotiations. It recognises that if you want to get to ‘behaviour behaviour change’ you need to start with ‘listening’ and ‘empathy’. change influence rapport empathy active listening 12
- 13. Clearly his computer isn’t listening to him. When interfaces don’t listen we get angry. 13
- 14. It looks like you’re giving a presentation Get help with giving the presentation Just give the presentation without help. Don’t show this tip Clippy has all the elements that we’re supposed to include in emotional design. He’s again informal yet direct. He’s cute looking. I’ve always thought there was a lot of good thinking behind him. But people hated him. Why? Because he’s so bad at listening. 14
- 15. My hero Clifford Nass redesigned Clippy very simply to listen and empathise. When Clippy offered advice, he would ask ‘was that useful?’. If people said ‘no’ Clippy would say ‘that really ticks me off. Let’s tell the folks at Microsoft I need to be reprogrammed.’ and Clippy would encourage them to write in and vent their dissatisfaction. And it worked: users Clifford Nass 15
- 16. When you watch this video, you notice that there are lots of signs that things are going wrong. Our guy begins by giving the screen a hard stare. Then he seems to swear. He slaps the keyboard. And then he really starts to lose it. A more sophisticated computer might have picked up on those warning signs and saved itself a beating. 16
- 17. In Affective Computing, Rosalind Pickard suggests that computers could use input from many sensors (facial recognition, audio input and so on) and pattern matching to detect users’ emotions. She recognises this is complex (even people occasionally misread emotions). We’re a few years away from this. But Clifford Nass shows us that listening can be as simple as saying ‘how am I doing?’. 17
- 18. The FBI knows there’s a right way to empathise. Don’t say ‘I know how you feel’. It’s too easy for other person to say ‘Oh no you don’t’. Wrong: Instead, show you care and create opportunities for dialogue. You don’t need sophisticated technology for this. In fact, ‘I know how you feel’ it’s been around since the 1960s. Right: ‘I’ve never been in your situation before, but I imagine you must be feeling very depressed and lonely’ 18
- 19. Eliza is a computer ‘therapist’ that asks users how they feel and uses pattern matching to respond and draw them out. ELIZA It’s crude, but good enough to pass a basic Turing test - some people think Eliza’s answers come from a real person. 19
- 20. FBI negotiators know they need to project the right Be positive, upbeat personality. Reassure hostage-taker that Now imagine if you hooked up Eliza’s pattern matching to an online database and things will work out well gave it an upbeat personality with a bit of edge to it. Be credible That sounds familiar... Show you understand their reasons but don’t be too eager to please 20
- 21. Siri is an evolution of Eliza’s pattern matching approach but with better jokes. That creates a personality and a basis for empathy. 21
- 22. Clifford Nass ran an experiment where he gave participants blue wrist bands and asked them to complete tasks a computer. For half the participants he put ablue border on the computer screen and said ‘you and the computer arethe blue team’. For the other half, he gave the computer a green borderand said ‘you’re the blue guy working on the green computer’. When the colours matched, people tried harder and thought the computer was smarter. Building rapport doesn’t require complex technology. Just good psychology. 22
- 23. Excellent. We’re half way up the FBI’s behavioural change model and we’ve not had to build an artificial brain. Listen, empathise, build rapport 23
- 24. The FBI has a lot to tell us about how to handle emotions that arise from situation. But sometimes conflict arises from personality differences. 24
- 25. For managing relationships, this book was recommended to me. I love it because it centres on a simple model. (Which we can use when we’re designing interactions.) 25
- 26. You have to understand Task focus people’s disposition (passive - aggressive) and motivation (task - relationship). Passive Aggressive Relationship focus 26
- 27. In the centre is the ‘normal zone’. At the Task focus edges of the graph are extreme types who can be difficult to get along with. Whiner Tank Passive Aggressive Yes person Think they know it all Relationship focus 27
- 28. What I find normal and acceptable will Task focus be different from what you find normal and acceptable. Everyone is someone’s difficult person. Whiner Tank We need to tune our behaviour and responses to get the best out of those conflicts. Passive Aggressive Yes person Think they know it all Relationship focus 28
- 29. And the secret to that is understanding Task focus the intention that drives those personality types. Get it right Get it done Passive Aggressive Get along Get appreciated Relationship focus 29
- 30. Maja Mataric has built socially assistive robots that coach stroke victims through their physiotherapy. She tried tuning the coaching to each patient’s personality. The robots would tell extroverts ‘Come on, try harder’. But introverts would be told ‘I know it’s hard, but it’s for your own good’. And... it worked! Patients preferred the robots that were tuned to their personality and tried harder for them. Maja Mataric 30
- 31. If you need to be appreciated, you’ll like the fact that TripAdvisor tells you when people have read your reviews. If you’re a ‘get it done’ kind of person, you might feel this was unnecessary and spammy. 31
- 32. And if you’re the kind of person who needs to get along, you’d appreciate this error message that says ‘It’s my fault’. (Personally, I find it rather craven.) 32
- 33. You don’t need to give your users a personality test before they start. You could learn their personality in the same way that Pandora learns your taste in music. 33
- 34. And you might even be able to pick up information about users’ personality from specialist services. 34
- 35. Task focus So instead of designing fixed patterns of behaviour, maybe we should design flexible patterns that adjust to users’ disposition. Get it right Get it done Passive Aggressive Get along Get appreciated Relationship focus 35
- 36. Behavioural Change Stairway Model And we’ve seen how a model like this can help us think about behaviour listening, empathising and building rapport, rather than change rushing towards outcomes. influence rapport empathy active listening 36
- 37. A lot of the discussion of emotional design today centres around ideas of ‘brand’. And on old-fashioned, static, monolithic brands. But I hope I’ve shown you that we can do much more. We can create a flow of emotions and more dynamic, adaptable personalities for our designs. We’re designing responsive web layouts, why not responsive interaction rules? http://en.wikipedia.org/wiki/File:Coca_Cola_ad_ca._1943_IMG_3744.JPG 37
- 38. If we put some emotional intelligence into the interactions we design, we can disarm conflict, create richer, more memorable experiences, and improve performance for our users. The future may bring us better ways of reading emotions. But the tools we need to get started are already in our hands. Let’s use them. http://www.instructables.com/files/orig/F4U/KUUR/FGVL6H7X/F4UKUURFGVL6H7X.jpg 38
- 39. What are emotions and how Still the go-to text on using emotions Applying the triune brain model How law enforcement agencies could we create them in a in computing. to the problem of creating negotiate in highly charged, machine? emotionally resonant designs. criminal situations. A guide to recognising A refutation of the idea that Original research into how people Applying theories of story to emotionally charged relationships emotions have no place in higher make ‘gut’ decisions. interaction design and to getting the best from thought. them. 39