Hello and welcome back. Today we're going to talk about high-level models in human and computer interaction, and there's three theories that we're going to get to. We're going to talk a little bit about distributed cognition, activity theory, and situated action. And, as before, I can really only give you a brief overview of these theories, as they're actually quite complex, and so I want to start in something more general. So what does it mean for a theory to be valuable? At this point in the course you've probably heard about some low level theories like Fitt's Law. And that theory really makes sense. It's about how fast somebody can click between two points. It's very measurable, it's very predictive. And a lot of people find it to be valuable. But,I read a paper when I was first studying high level models that talked about kind of what theories can contribute to a field that went beyond just this idea of being predictive. And so I include a citation to that paper here, and if you have access to the digital library, you can follow along. But I'll cover the main principles from it. So, the main tenet of the paper was that the value of the theory is not whether it provides an objective representation of reality, but rather how well it can shape the object of study, highlighting relevant issues. So in complex high level theories, the goal is not to be able to perfectly predict what's going to happen. Usually these theories address settings and contexts that are so complex that it's very unlikely that you're going to be able to predict things directly. But rather does it let you focus on issues that are most relevant? So in the paper they discuss four powers of HCI theories, four things that make HCI theories good. And these are descriptive, rhetorical, inferential, and application powers. So let's go through each of these in more detail. Descriptive power is whether the theory provides with you a conceptual framework to make sense of the world. So does it take a really complex situation and make it make more sense without maybe over simplifying things that are really important to understanding them. Rhetorical power is about the ability to describe the situation both to yourself and to others. So some of this is kind of the idea of language, like how do we term these complex interactions that are happening in social settings, for example? Inferential power is about the ability to make inferences or predictions about phenomena and interventions. This is kind of what I describe with Fitt's Law. It's very good at making inferences. If you tell me how big two targets are and how far apart they are, I will frequently be able to make a pretty good prediction about whether a person can click faster between setting A or faster between setting B. But that's not necessarily the case with high level theories. And lastly there's application power. So can applying this theory inform and guide system design in some sort of a way? And so usually at this stage I actually ask my students to reflect on whether Fitts' Law measures up using these attributes. And I'm always kind of hesitant about this law, because I think it's so low level that frequently it doesn't provide a whole lot of value for designing complex systems. And students knowing that I'm very hesitant about Fitts' law try to be very critical towards it as they evaluate it. But in fact, I surprise them by telling them it actually does pretty well on all of these criteria. It does provide a very good conceptual framework to make sense of very specific interaction, which is how people how fast people click between two points. It does let us describe those this ourselves and to others. So for example, if we describe the size of a target and how far away it is from another target, we've actually described the entire situation. It lets us make inferences about how fast people are going to click. And it lets us apply those to informed design, so for example, putting menus at the top of the screen, so that essentially becomes an infinitely sized target. Other theories are not necessarily quite this straightforward. So three high-level theories we'll talk about today are distributed cognition, activity theory, and situated action. So, let's start with distributed cognition. And those in the know call it DCog. So if you want to call it DCog, you'll be one of the cool kids. So some key tenets of distributed cognition is that action is coordinated among embodied agents, where an agent could person or an agent could actually be a device. So for example, if I am making a shopping list, I'm actually coordinating actions between two embodied agents. Myself, I'm the embodied agent who has some ideas of what to buy in the store, and my notepad, which is actually going to store those ideas for me to reference later. Both are actually embodies agents. So, in essence, I'm kind of delegating the work of remembering to my notepad. Information itself can be embedded in an object or in an action. So, for example, in that notepad, I have embodied the information of the things I need to buy in the store. And all of this is part of a kind of a large cognitive ecosystem. All the different agents who work with a shopping list, maybe I give the list to my husband to go buy things at the store. We've added another agent to the system. All the devices that are used along the way, so maybe he then decides to do some comparison shopping using his phone. So now the phone is also an embodied agent that's part of this ecosystem. All of this is part of a single system, and you have to understand how it all works together in order to understand the entire system. So this was a theory that was originally developed by Edwin Hutchins. And he actually originally developed it by looking at how pilot, copilot, and the devices in an airplane cockpit all work together. So he looked up this idea of speed-bugs which were little plastic bits that the pilot could place on the speed dial of the airplane as you see in the photograph. And they all signified special things, such as, when the arrow is close to this speed-bug, that means I need to decelerate or I need to open the flaps of the airplane. I personally don't know how to fly and airplane so I'm probably saying completely ridiculous things about it. But it became kind of a visual indicator. So no longer did the pilot need to remember exactly at 150 is when we need to decelerate. But rather that they could just compare whether the arrow was close to the bug or not close to the bug yet. And so, in some ways that process of cognition, the process of remembering the correct speeds at which to do certain actions was actually kind of embodied in the actual physical devices that were part of the cockpit. So when doing distributed cognition, I think the most important thing is actually to ask yourself, what is the unit of analysis in this cockpit, or in this context? So what Edwin Hutchins found is that in the context of an airplane cockpit that had a pilot and a copilot, the unit of analysis that was meaningful was actually the entire cockpit. So if you just observed the actions of the copilot or just observed the actions of the pilot and didn't have anything, any context about the kinds of technologies they used, you would of have an incomplete picture of the kinds of actions that they took and the challenges that they faced in flying the airplane. Activity theory sort of extends this idea of what is a meaningful unit. And it actually posits that maybe the entire activity, so the process of acting upon an object to transform it into an outcome, maybe that can actually be the unit of analysis of a system. Activity theory really, one of the key tenets is the continuous change and development of systems. Systems don't stay the same. And lastly, it's the fact that artifacts such as tools, rules, social structures around an activity all mediate how an activity is done and specifics of that activity. So, activity theory is frequently represented as this kind of a triangle. So if you look on this triangle, you see the subject that is the person doing an action. There is kind of a line, a light line connecting it to the object. You can think of the object as the objective or the thing that's being transformed into an outcome. So it's kind of the goal of the activity. And if you notice there's kind of a light line, or a dashed line, between the subject and the object. Because if there's a tool that's required in order to manipulate the object, then that tool mediates that interaction. But activity theory also recognizes that people very rarely act in isolation. People are frequently parts of communities, part of a social context. And so one of the things that you notice is that the community can also affect the object. The community typically affects the object through a process of division of labor. So every person who's kind of part of the community may have some responsibility for some part of the goal, some part of the object, and that mediates the interaction with it. And lastly, the subject, the person who's doing the action, is not just kind of part of the community outside of any sort of context. But rather their interaction within the community is mediated by the rules of that community. And so this kind of captures the mean principle. Though lots of people say that actually all aspects of this model are connected to all other aspects potentially. But typically this is the way you see it represented, as a triangle. So let me give you a concrete example from my work. So I was looking at how parents raise children in divorced families, and actually it made a lot of sense to talk about it in terms of activity theory. So if you think about the objective, the goal of the system as raising a child to become a successful adult, there could be a number of subjects. So the remote parent, the local parent, the child themselves, they all contribute to this objective. Now in the process they're actually part of the community. So the child may have two different families that they're part of, their mom's family and their dad's family. They also are part of a larger community, for example, the school. They may have extended family that are also part of this community. And the way that the parents and the child interact with the community is through rules. So that may be specific household arrangements, custody arrangements, cultural conventions. All of these are part of it. Now when you're apart from the child, frequently the way that a parent, a remote parent would interact with a child is very much mediated by tools. Things like the telephone, or Skype, or sending a letter, these are all examples of tools. And there's also this idea that if the whole community is responsible for raising a child in one way or another, there's some division of labor. What is the child responsible for in their own upbringing? What are the parents responsible for? What are the teachers responsible for? And so, actually kind of classifying the problem in this way made me think about all the aspects of it, perhaps aspects of it that I wouldn't have thought of. So, for example, maybe I wouldn't have originally included the consideration for the extended family or the consideration for the child's school as part of my analysis. But by classifying it as part of the community, I was forced to think about these issues. So a little bit of kind of review and also talking about the relative strengths of distributed cognitions and activity theory. So, distributed cognition really has a very flexible unit of analysis. So, for example, I'm in a studio right now. And I may decide that the unit of analysis here involves all the technology in the studio, me, and the cameraman. We're all part of a single unit of analysis. Or I may decide that maybe the unit of analysis is better set as the MOOC. And all of the different professors who are involved in it are part of the system. One interesting part of the distributed cognition is that it really focuses on kind of exposing system workings. So small things like how do we use the tools to achieve what we're trying to achieve, how it all contributes to the larger system, the larger ecosystem. And many people find it to be more easily applicable to design because you can clearly see the role of specific technologies in the process. In contrast, many people find that activity theory constructs the names better. So it's easier to actually refer to different aspects of the system. So if you want to talk about the community, it's maybe more difficult to do that in a context of distributed cognition, because you have to draw up new boundaries. The other thing is activity theory really emphasizes individual agency. So while in distributed cognition both I and the notepad on which I write my shopping list are actually sort of equivalent, we're both agents in the system, activity theory doesn't think that way. The notepad becomes the tool, but I, myself, am the subject. I have agency, I have motivations. I have goals that I'm trying to accomplish, and the notepad does not. So it kind of separates this idea that people and technology are different. And the other thing is activity theory really thinks about this idea of process and really forces you to consider how that system is going to change over time. And distributed cognition is really more about capturing something at a particular snapshot at a particular moment in time. But I also like to say that the two are actually quite similar, so that both can be applied in a wide variety of situations. They're both particularly relevant to social and collaborative computing, where you're not just dealing with a single subject, single person, single agent, but rather with a whole system where there may be multiple people interacting with something. They both try to incorporate this idea of social, cultural, and cognitive context. The idea that we're not just doing our action in the lab, with nothing else around us and nothing else influencing us, but rather that we're all doing it as part of a larger system. And actually, the methods used by both of these are actually fairly similar. Frequently, it's ethnographic data collection. So really getting out on the field, observation, contextual inquiry, really asking people questions to figure out what they're doing, and observing them to see how they go about the steps of achieving their goals. And so now I want to talk just a tad about situated action, and this is really a theory that I feel like I can't quiet cover in the time that I have. And so, if this sounds interesting to you, if you have kind of questions based on what I say, I know I did have a lot of questions when I first read about situated action. I really encourage you to read Lucy Suchman's book, Plans and Situated Actions. And as you see, the subtitle of it is, The problem of human machine communication. So it is actually immediately applied to technology and how technology affects people. And the main tenet of Situated Action is the idea that human beings are master improvisers and explainers. So frequently we like to think of ourselves as having a plan, and then doing our actions one after the other in order to achieve the plan. And this book uses ethnographic data to really challenge that assumption. It points to the fact that humans just do. And only when they're asked to explain what they're doing, did they retroactively come up with a plan that would have accounted for their actions. So kind of a few basic tenets that go along with this, so plans are representations of situated actions. So actions are situated in the sense that you kind of are doing your best in any moment to do the next right thing. Whereas a plan is just something that maybe you use as a resource or use to represent whatever you did after the fact. In fact, these representations, like actual step by step plans, typically only are witnessed and occur when an activity becomes problematic. So for example, if you were riding a bike, you're not really planning on how to ride a bike. You're not planning to put your left foot on the pedal and push, put your right foot on the pedal and push. You're just doing it. But if I stop you and I ask you hey, can you teach me how to ride a bike? Well in fact, I actually don't know how to ride a bike so this would be a useful interaction. If I stop you and ask you how you ride the bike, then you're actually forced to consider and think about how you would plan on going about the process. Well, you need to gather some speed so that you don't actually fall over when you go on the bike. And here's how you turn. You have to kind of lean your body into it. But you only come up with that plan and that explanation of it when I make it problematic, when I ask you to account for it, when I ask you to explain what's going on. The idea of objectivity, so kind of the objective right way to ride a bike, or mutual intelligibility, which means how do you explain something in a way that both people can understand it, is not a given, but rather something that is achieved through hard work. We both have to work together pretty hard to make sure that you're explanation of how to ride a bike is intelligible to me, and for you to understand that my understanding of it is actually what you meant to communicate. And language is seen a central resource for achieving that objectivity, which is why it can be so important to observe two people who are interacting together around a technology and observe the kind of things they say to each other through the process. Just kind of the quote that I really like about situated action is that a basic research goal for studies of situated action is to explicate the relationship between structures of action and the resources and constraints afforded by the physical and social circumstances. So it's this idea of, how do you explain actions as they occur, given the specific constraints around the social situation of the participant or the kinds of physical tools that they have around them in order to achieve the action? Now, this may sound just really hand-wavy. These are very high-level models, and I don't expect you to really know each model inside and out from a ten minute lecture. What I would encourage you in doing is actually reading more about these models. I'll give you some resources at the end of these slides. But also knowing that there are a lot of these high level models that work in certain contexts and are good at explaining certain parts of the problem. That have high rhetorical, high descriptive power, maybe high application power, but are not so much good at predicting what a person is going to do. There are, in fact, also other theories that are quite relevant to HCI, that may be relevant depending on the context you're working in. And these I'm not going to cover at all, but basically just to give you ideas of what some of them are called, names of terms. The actor-network theory is a really popular one that treats sort of people and objects all as part of a single network. If you're doing anything that involves people to learn, it maybe useful to you to look at a few learning science theories like constructionism and constructivism. If you are designing a system where people are exchanging social capital or goods, economic theories may be really useful to you. Of course psychology theories are I think useful to every HCI user interface designer, and we do cover quite a few of these in this course. But also if you're working with a specific age, so whether it's children, elderly, adolescents, it's useful to also maybe take a look at maybe some human development theories to know what is developmentally appropriate for a particular stage in life. So as I promised, here's a few citations for more information. So for situated action, I really point you to the book by Lucy Suchman. I think it is impossible to explain it well in five minutes. But hopefully you're at least familiar with the terms enough to know when they may be applicable and to know if you're confused. And the other paper is the paper that I started this lecture with, the paper that considers the different powers of theories. And in fact it actually talks about distributed cognition and activity theory in terms of these four principles. So this role of theory in technology design paper by Halverson I think is a really good read and a fairly straightforward read so I really do recommend it. So thank you for joining today. I know this is quite a different set of theories than ones we've covered in this course before, but hopefully you learned something new. And I'll see you in the next video.
