SINFO is a well know event amongst computer engineering and computer science students in Portugal. The 24th edition took place at Técnico Lisboa (Alameda campus) between the 20th and the 24th of february, and gathers, beside the obvious target audience (students and technology enthusiasts), interesting and influential people and employers. For us, the people we have to pretend we are most excited about every year are the people that are somehow involved with the videogame industry: previous editions brought people from Naughty Dog, EA Games, Bioware, Blizzard and much more. As for this year, we got Leonard Boyarsky from Obsidian and Mike Ambinder from Valve.

Mike Ambinder calls himself a experimental psychologist. Studied Computer Science and Psychology at Yale, PhD in Psychology at Illinois. Started working at Valve and helped out on projects such as Team Fortress 2, Left 4 Dead, Alien Swarm and Portal 2, focusing his work on user experience and adaptive systems. He’s one of the pioneers in applying psychophysiological techniques in game design. It’s a sunny but windy day, 11:50AM and temperature is close to 16ºC (more or less 60ºF) and after abusing on SINFO’s coffee-break pastry (and for that reason showing up late), I got the chance to interview Mike.

Rubber Chicken: Is this your first time in Portugal? Are you enjoying your stay? What do you think of our country?

Mike Ambinder: Yeah, its my first time in Portugal, still trying to get a good feel for it but yeah, always hear good things about Lisbon and now I’m happy to finally be here to check it out, I’ve had great portuguese food last night, I can’t tell you the names of the dishes but we had a bunch of those. I’ve been told I should check downtown and the old historical part. Everyone’s been incredibly nice, kind and generous and has been great!

Rossio, on Lisbon’s downtown. Also has a lot of PokeStops.

RC: Can you describe what you’ve dined last night?

MA: So I had a couple of fish dishes and some pork dishes. I can talk to the people who ordered for me, I’ve just trusted them, I told them “Get me great portuguese dishes”.

RC: But did you liked it?

MA: Yeah yeah yeah, it was great! I wanted to try traditional portuguese food and they definitely hooked me up, it was great.

Cozido à portuguesa, a mixture of boiled portuguese meat, sausages, vegetables and rice.

RC: SINFO is an event to allow attendance to interact and be closer to influential, important and interesting people or companies in the fields of information technology and computer science. What are your thoughts on this initiative and what do you think you can add with your interaction with general attendance and students?

MA: So, the idea is awesome. Trying to get people who are starting out their careers, getting them exposure to people who are in their careers, you can have practical advice, you can talk about challenges and problems they face, and how can you prepare yourself as a student to actually enter this fields and do well. So the more opportunities students have to do this, the better they will be. It’s kinda hard to think about and envision things in the abstract, so getting some concrete examples, for instance, if you want to be a developer at a mobile apps company, well then you should talk with people that develop mobile apps, that understand the kinds of problems they have and what kind of skills they value. Then you can best position yourself to go and do that. So, I’m a psychologist at a videogame company, it’s not a common job. When I was starting out, there was really nobody for me to talk to. I had to carve my own way, I didn’t have guidance or people that were senior in the field and that could point me in the right direction. Some people are doing stuff, but weren’t many resources out there. I remember how it was like when I was trying to start out, now that I’m more established on the field and have been around for a while I can help people to point to the right direction and I’m happy to do that because I understand how difficult it was not having people around previously. So, to come to places like SINFO where you can interact with people who are doing the kind of things that hopefully you can do someday it’s pretty amazing. I can, well, who knows how helpful I’ll actually be but you know, give practical advice and practical examples where I can help people learn how we think about problems and the game industry from a perspective of a psychologist and just in general, any general advice I have for students to enter the industry I’d be happy to give.

RC: You describe yourself as an experimental psychologist. Can you give some insight about your work at Valve and what do you enjoy doing most in projects?

MA: So yeah, I call myself an experimental psychologist, so broadly speaking what I do is looking at the application of knowledge and methodologies and how can we apply them to game design. Where knowledge and human behaviour can be useful, you know, there may be research in psychology about human behaviour that we can try and apply and make use of it on game design, or we can design experiments that would help us give us insight why people are doing what they’re doing. Some specific examples of that might be how to manipulate visual tension on the screen, how you can get people to look where we want them to look and not look where we don’t want them to look. Designing experiments for in-game economies and trying to build a knowledge base in how people interact with these things. How do you foster cooperation or competition amongst players, how to design reward schedules, like how often should you give somebody a reward and what kind of reward, that kind of things. Questions like that! Psychologists study human behaviour for hundred years and there’s a lot of knowledge built about what we do and why we do it, and so, trying to make use of it in game design context it’s what I do. I also end up doing a lot of data analysis, you know, I’m a stats nerd, so I spend a lot of time with gameplay data, looking up data on Steam, a little bit of hardware research. I work in play testing methodologies, what methodology is best suited to gather data from players, eye tracking vs surveys vs Q.A. vs biofeedback things vs, you know, there’s a variety of mediums that can give data to you, it’s about what methods can we use to get the data we want. But in general, these days, it’s mostly data analysis and then supplying psychology to questions of interest.

Mike Ambinder

RC: How do you deal with a conclusion after looking up stats and discover that actually doesn’t apply?

MA: We just want to acquire information, we want to be smarter. Sometimes that information will be just “oh, this works”. Sometimes will be “oh, that didn’t work”. So, in either case we got smarter. So we’re happy to try and fail as long as we’re failing intelligently and acquire information, the outcome is less important than the question we ask, where you can get some more insight in what is working and not working, cause both are valuable.

RC: So, i guess is more of an iterative process?

MA: Yeah, yeah, that’s a great word to describe it. It’s a constant process of hypothesising, gathering data and then iterating and improving that hypothesis based on the data.

RC: Portal 2 is the only not multiplayer focused game that you’ve worked on? 

MA: Yep.

RC: So since it’s a puzzle game and a genre that’s unusual to have multiplayer, can you describe the greater challenges you’ve faced during development on the multiplayer aspect of the game?

MA: So if you’ve played single player Portal you have a basic understanding of how portals work, but, the challenge of multiplayer was trying to figure out how can you take what used to be a single player experience and actually make it enjoyable and comprehensible for two people to playing it at once, how do you build a cooperative version of that, and there was a lot of iteration. We ended on something that was pretty good, but it took a lot of work to get there. So the question becomes “How do you design puzzles that require the use of potentially four portals placed in different ways?” and “Are there enough clues in the environment to let the player know exactly how to build up a solution?”, because, you know, placing two portals anywhere is bad enough right? So placing four, and then having to figure out “step one interacts with step two interacts with…” that’s complicated. A lot of that was environmental design, making sure you give players clues in the environment so they can figure out. Here’s a potential opportunity to place a portal here so what’s the consequence of that?

But you wanted to be a collaborative experience so people are solving it together, you don’t want people to work separately on their own, they could but if they act like “Well, if we start here, then the player will come out of this portal and land here but we want to land up there, so how do we do that?”, that kind of thing. It was a challenge how to enable or foster that collaboration, a lot was iteration, it was just trying out different things and trying to come up with puzzles that weren’t overly complex but were challenging enough.

Best friends forever!

RC: So a lot less on the theoretical side, I guess?

MA: Yeah yeah, it’s not like we started knowing what was the structure and frame of what we want to apply, it was a language we had to create to ourselves.

RC: Do you have any advice to someone trying to enter the videogame industry?

MA: The single most important thing you can do, whatever your discipline (so if you’re a psychologist, a computer scientist, a level designer, artist or modeller), is: create anything and then place it in front of people that are not you. Gather their feedback. Then iterate and improve whatever it is you’ve created. That’s how we make games, that’s Valve philosophy of game design. We have an hypothesis, things we want to try, we make it, we put them in front of people and gather their feedback and then iterate and conclude the products. Thats why our product take forever to make, and if you like our games that’s why they’re good. And so, people who are good at that process and have done that process and shown that they’ve done that process, are the kind of people we look to hire. So, you can be 13 and making something or you can be in the industry for 15 years and making something. We just want to see people that have gone through that process of creating something, then iterating and improving based on feedback they’ve gotten from people that are not them. As a psychologist: we are very bad at evaluating our work objectively and so we want the greatest number of honest sources of data we can get.

So, for instance, if you’re an artist and you’re asked to draw a cool barbarian, understand why you think it’s a cool barbarian and what your goals are and then ask other people and think how they react to it and improve your barbarian, make a better barbarian. That’s how you get good at whatever part of making games you’re into. That’s the best advice I can give, whatever you want to do just make something and improve it upon feedback of persons that are not you.

Mike Ambinder

RC: So, I have one more question based on curiosity: I don’t know if you know The Witness (MA: yes I do), but Jonathan Blow (the lead designer of that game) doesn’t believe in that iteration of process of play testing. So, since you are a psychologist and tend to see things from a more theoretical side, what advantages and disadvantages do you think each of this methods have and how do they adapt more or adapt less to what are you trying to achieve?

MA: So, everybody has different perspectives, by no means we think that one way of doing things is better or superior than the other. We just found one that works. From our perspective we’re a bunch of really smart, some of the best game designers in the world, right? But we can never be certain that we have a good idea. We can think we have a good idea and we can be pretty confident we have a good idea, but we don’t really know until we actually test it. So we like to gather data and essentially just validate our hypothesis. We don’t want to just assume that we’re making the right choice as best we can, there are always constraints and we can’t always test everything (of course), but in general we want to try and be as smart and confident about our decisions as much we can. And the best way we found to do that is to have an hypothesis, create the game design and gather data from people that are not us.

Don’t press it, don’t *aaaaAaAaAaAaHHHhhh*!!!

Different people work better in a variety of ways, Jonathan Blow obviously has created amazing games (that are awesome!), but his process is more top-down driven right, where is “hey, I have an idea and I think it’s really cool, lets implement that”. And you know, it’s a single artistic vision, that great! We just want the best ideas to come out, well, we don’t want the best ideas and just assert it’s the best idea, we want to understand why that idea is the best and see if the data supports that. It’s not that one way is better than the other, it’s just better for us.


RC: And when do you know a product is ready for shipment?

MA: Yeah, it’s tough, there’s a bunch of people that’ve been around for a while, that released a bunch of awesome games, so they have pretty good sense and go “so, the feedback we have from people it’s pretty good and of we go”. But it is one reason why our games take so long to make, because we try something and we don’t want to ship it because there’s a artificial deadline, we want to ship it when we think it’s ready. And so it really comes down to a group of senses, like “hey, we’re pretty confident in the mechanics, we’re getting good feedback from data and we think we’re ready to go” and also, the way we make our games now, when we ship it it’s the beginning of the conversation with the community. All of the sudden we can get real data, data from millions of people! Shipping the game is not the end, it is really just the beginning.

RC: Thank you very much for your time Mike!

MA: Thank you, nice to meet you!