 Two people you have met in your lives, who are both intelligent, intelligent as per you. An intelligent person that you dislike, it could be anyone, it could be your boss. Bosses are always intelligent, many people dislike them. And it could be your colleague, it could be anyone. And in this box, I would like you to imagine an intelligent person whom you've liked. Again, it could be anyone, it could be a friend, your spouse or someone you've met. And I would like you to spend a few moments thinking about why this is the case. They were both intelligent but you liked one person and you disliked one person. What were those things because of which you put that person in that box and this person in this box? As you think about this, I'm going to introduce you to Cogbooks. Now Cogbooks is an adaptive e-learning company that makes online courses. Now all of you must have gone through online courses like Udacity, Coursera, which is nothing but a sequence of pages. What Cogbooks does is it analyzes how each student is learning and based on that, customizes the course for each student so that every student becomes successful. This is done using machine learning to analyze how the students learn and then recommending the right activities for that student to learn. So if some of you who is a senior person and some of you who is a junior person both take on the same course, both are not given the same activities to do. So that is about Cogbooks and this is what happened at Cogbooks. We built an AI-based product and we thought, okay, Amazon uses AI, YouTube uses AI to recommend videos to people. It works for them. Amazon uses AI to recommend products to people. It works for them. So what we should do? We should use AI to recommend learning activities to students and it will work for them. It didn't work. Nobody liked our product. So when we had a product which shows learning activities and some recommended more learning activities for you because this worked for people like you, nobody visited them and it didn't really help them succeed. So even though those recommendations were there, they were not having any impact and we didn't get any positive feedback. And why was this happening? For two reasons. One, it felt mechanical. A learner is going through a course and we are just saying, okay, these are the recommendations for you. It works for things like YouTube where the content, their attention is on the content and their emotional needs are different. But it doesn't work for an online learning environment. The second part was it lacked empathy. So the learners were going through and we were showing recommendations, but there was nothing we were doing to handle their emotions. That means it lacked empathy. And the reason that happened is because we used the old design process. I'm assuming most of you guys are designers. So I want to go into what personas are. But we defined personas. Then we defined some common scenarios that these personas go through. And then we designed an interface that works for most of these scenarios. That's the usual design process. So we followed it for this also. But then it did not work. And then we got into finding out why it did not work. And it turns out that people think of products as people. And this was a fundamental shift for us. And an extension of that is intelligent products are thought of as intelligent people and not just products with some AI based component in them. So how do we make sure that we design our product that is perceived as an intelligent person, not just as a product with some recommendations in it. And we just saw in the previous slide where you imagined people that just because someone is intelligent doesn't mean they are likeable because of various reasons. So I'm coming back to this slide where you had an intelligent person you disliked and an intelligent person you liked. The reasons why you disliked that person and the reasons why you liked that person does anyone want to share any one reason? You won't have to reveal the person. So it's not invading your privacy. Anyone wants to share why they disliked an intelligent person in their lives? So probably dislike will be with confrontations. Too many to-and-for discussions. And liking is mutually agreement on the ideas. So an intelligent person, although dislike is a strong word but I usually sort of did not get along very well was someone who would always try to force his way or is always of the notion that either it's my way or the highway. And an intelligent person that I liked is someone who is very open, who is very transparent and who is at least giving you a patient hearing, trying to understand your perspective and comes to a conversation without a prejudice or a bias. Thank you. Both fantastic answers. Thanks a lot. These are exactly the kind of insights that we got when we studied intelligent, likeable people and intelligent people whom we disliked. So like what these members shared with us, the intelligent people whom we like try to understand us and make their intelligence work for us and the intelligent people whom we dislike are intelligent by themselves but they don't do any effort to make their intelligent work for us. That is one of the reasons. And in some cases in which they shared they even make them work against us so it turns into a confrontation. So how do we make sure that since our product is being treated as a person, it is treated like this person. It's perceived like this person and not that person or anywhere in between. And we got proof for this in sci-fi films where AI is almost always shown as a person and in this specific film X Machina, anyone seen X Machina? Okay. It's the same AI person and we've shown how that person can be likeable at some point and evil at another point. So all these things kind of came together and we thought what is the process that we should use to make our product like a likeable person. There were two approaches. The first approach is we can make the product having universally likeable traits. We can make it cute, have some cute drawings and everyone likes cute things but that only works for certain kinds of products and not all kinds of products. If it's a story, we can make it an underdog and we can make innocent attributes. So this was not the path that we wanted to take. The other path is that we can give them a positive personality, a personality that goes with the product. So if it's an online learning environment, the users don't expect it to be cute. What do they expect it to be? What do they expect in that environment? We thought to find out. And second one is about how do we understand the user's situations and give empathy to the artificial intelligence so that it treats each situation with empathy. So we defined a user persona, everyone knows personas and here I'm only showing the aspects of the persona that are relevant for this talk. So our persona was a student in US. It's a US-based company and he's a 19-year-old college student. He's studying biology and he's a low-academic performer and he's at the risk of dropping out of college. So what do we know from this? That we should treat him with care and in an online learning environment, if we treat this guy and say some strong words to him, it's likely that he's going to drop out of college and fail in his career. The next part was defining a brand personality or rather a product personality. So we are all familiar with user persona but the same process can be applied to a product also. That was the new thing here. Now how do we define this? Now there are already books like Nicely Said and Designing for the Digital Age that talk about how we define a personality but both those books are talking for different reasons. Nicely Said is talking for voice and tone, Design for Digital Age is talking about visuals. Now how can we extend that process of finding a product persona to the interactions as well? So for that we held meetings with the CEO, the founders of the company and asked them if this product, if this environment was a person, what kind of a person do you think it would be? Would it be a funny person? Would it be a friendly person? Would it be someone who helps you? Would it be someone who is interested in taking you further? And we collected all these answers and just like how we build a user persona from user research we built a product persona from founders research. So now let's look at how can the AI with this product persona understand Jacob's situation, Jacob is the user and give recommendations with empathy. So like I said, this was the old method where we define persona, we define scenarios and we design an interface that works for all of them. This is the new method. We define the brand persona or the product persona, we define the user persona that we do once and then for every situation that the user faces in his experience, we understand what are the emotions that are experienced by that user and then we imagine a response from the brand persona to the user persona and we translate that response into a user interface. So imagine a situation where Jacob the user is answering an exam or it may not be a final exam of the year, some test and his teacher is there and he is not able to write that exam. What would the brand persona who is a friendly person do in that situation and how can we translate that action, empathetic action into a user interface was the thing we needed to do for each. So let's look at an example, just what I told you. Jacob answered an assessment. He needs to complete the current module by the end of this week, so time is limited and he has, if he doesn't do that he does not pass in the semester and he fails that exam. Now in this situation, what are the feelings, what are the emotions that Jacob is experiencing? One is the experience of stress. He's feeling bad that he couldn't pass the test. First only he's a low academic performer then he passed and failed a test. Then he's feeling insecure about his studies. Like most people who are bad at studies do. I failed a test, probably I'm not good at biology. I will quit biology altogether. That's the other emotion and then the next emotion is not knowing what to do next. So I failed in this test, now what do I do? And imagine he's going through these emotions and we just show him recommendations for you. That's really un-empathetic, right? So what would David, Dave, our product persona do in this case? He will take this situation coolly. He will say if you are a person interacting with another person he will say okay, you didn't get the test right. There is no big deal. You can learn and you can come back and answer it. And then he will advise him what to study so that you can answer this test well and do well in general. So it seems obvious when it's a human-to-human interaction but we miss out on these things if we design for a common scenario where all interactions are clubbed together in one or two scenarios. So what we realize is that every situation like this has to be listed out in a table. We made a huge Excel sheet on paper first. We had to list down all the emotions that the user will likely feel at this point and imagine the response of our product persona to those emotions. And then how do we translate this into a UI? That's a craft that we are still coming up with. So the first one, how does the user know that the product is taking it coolly? One sentence. Then how do we encourage him to study more? Again, another some UI element that encourages Jacob to study more. And then what should Jacob do next? That is the next steps for him shown in the UI. So once we show these elements in the UI that was nothing but a translation of Dave's empathetic response. So I'll show you an example of the UI. This was a question, again, sorry for the American context. It's about Civil War in USA. And the question is about what was the primary goal following the Civil War. And he was given options. And Jacob did not know the answer to this question. And he answered, he gave a wrong answer to the typical e-learning program. Now you see here, we are not converting the page with all red things. You got it wrong. And that was a wrong answer. We are just giving a sentence that said looks like you didn't get that right. But don't worry, you can answer this again after going through the recommended items. And the recommended items are sequenced in a way that the next item for them is highlighted. Not only that, it's presented in a magazine-like layout where they are told about what they should read next and why they should read it. They should read it because you appear to need help with the post-Civil War goals of Abraham Lincoln. This is what you don't know. So you read this and you come back. So this was one page. We did this for every single situation, every single interaction. And what ended up is a user experience that is nothing but a translation of a product personality which was empathetic in nature. And again, this was the process that we used and every interaction, every situation was based on an imagined response of the product persona. The AI-based product became human. And after this redesign, the product became phenomenally successful for its target market. The user started giving positive feedback. You can check out the positive feedback on our website, cockbooks.com and how it has helped transform the lives of people. But behind all that is this process. Design the persona and brand once. And for every situation, treat the situation with empathy. Thanks a lot. I would like to take up questions from any of you. We have time for one question because suppose that we can take it offline. Thank you. You wanted to ask one thing. While going through this process, you considered how the interaction would be on chats, on voice. Anything that you considered talking to these guys, when it will be actually implemented, it will be a voice based or something chat based. So you can get the actual emotions. So the product that we were building was not our chat bot. If it were a chat bot, the output would have been sentences. But since the product, that's a limitation set by the industry and other things that are out of the hands of the user experience. Since the product was a user interface, we limited our scope to what are the elements that we can play with within that user interface like images, text, blocks of content, etc. And we translated only those elements into this. So voice and tone is a part of it. Only in form of written sentences. Yes, yes, yes. So those sentences, each of those sentences were made for that situation to address the emotions that the person was feeling for that.