|
Notes on an Interactive Story ----------- (RH/RZ: 3/4/04)
Narrative Façade excels at making me care about the characters. I really want to interact with Trip and Grace – even though they are damaged people on the brink of an explosive breakup. I’m curious. I have lots of little “oh no!” or “a-ha!” moments - and those happen because I'm paying attention, invested. This is exactly the kind of drama we get from other familiar sources (television serials, films, novels). We peek into a situation that is just dramatic enough to feel compelling... letting it play upon our compulsion to know more. I’ve been through similar, but not quite identical situations - and it still hooks me.
On a fundamental level, this stuff is the earmark of a compelling story. It struck me very squarely today that Interactive Fiction has the potential to compel attention because it has a writerly focus – something that games often do not.
Interaction Because I find the characters and situation of Façade compelling, I want to influence them. When Trip and Grace move around their yuppie apartment, offering me drinks or asking me to join them on the couch – I really do feel invited to participate. But my control is somewhat limited, as is the feedback about my actions. Throughout the play session I couldn’t shake the feeling that what was typed into the window had a glancing impact on the overall dialog. I felt excluded, at times, from the powerful and dramatic situation just there before my eyes. Partly this is the Natural Language problem. It still takes a little while for Grace and Trip to recognize and respond to typed words; the interaction sometimes feels more like a laggy IRC chat than a real-time conversation. But even stranger, improvements in graphics and sound made me feel less sure of myself. As the characters become more "realistic" in their response to one another, they feel further away from me. My casually typed dialog feels stilted and out of place. It’s hard to be as good at acting as they are! Rob and I discussed this for a long time. What was it about the interaction that made us feel excluded – despite our understanding of the code, our interest and overall intent? Consider the three participants:
Grace and Trip are on the same level – they interact with each other with complete clarity. They get each other. This is true from a character standpoint as well as a mechanical one. As you type into the window, the conversation often moves on without you (maybe you missed a queue, maybe you didn’t know what to say…). When this happens, their synchronicity is reflected back at you – a counter-example to your relationship to them as a couple. You really are an outsider. And in a way, that’s the whole point. The “game” is that you can develop a higher level of intimacy with one of them by choosing sides – or try to keep them together. While it’s interesting to try to nudge Grace towards a breakdown with repeated digs, it’s sometimes difficult to tell if your words are having any real effect. At those times, they seem intent to dialog - with or without you - and that hurts.
Ok, not really. But you notice it! A while back, when Façade was slower, it felt less like this. However, Grace and Trip felt less “real” then, too. They seemed to pause too often, like robots waiting for input. Rob also noted that for him, the inclusion of voice (instead of word baloons) changed his perception of them. With text – he felt like things were actively being generated – that his words were causing other words to pop up. With spoken dialog– which seems more “natural” – it felt less generative (and for him, less user-driven). I found myself asking – is this the result of a theatrical approach to the overall interaction? Does good dialog naturally exclude the (amateur) player? Moreover - would an action-based approach (the “simulationist” or “emergent” perspective) be at odds with the one taken here? Are we talking about the difference between telling a good story and designing a good game? Feedback Obviously, it’s not that simple. And obviously, I can’t believe that narrative and interaction are at odds – my entire motivation as a researcher has been to link these two things expressly via artifacts and theory. But it did set me back. More discussion revealed that the processes by which Grace and Trip interpret user action and then respond are complex – and usually pretty on point. But they’re not very transparent. It’s hard to know if they’re “getting you”. In CS speak: it's hard to know what the state space is at any given point in the narrative. For me, the human user, it feels as big as my imagination and vocabulary combined. This leads to the classic Interactive Fiction problem – the actual space of possible interactions is quite small (at least by comparison). I can type a lot before striking upon something that "works". I can string such phrases together to create changes in state - but this doesn’t feel very “natural”.
It's important to give the user plenty of feedback about what they are doing... so they don't get stuck or frustrated in this way. But it's hard to do this without jumbling up the dramatic dialog - especially when the input is text-based. Interface In the end, I decided that I don’t really care about having a natural conversation with Trip and Grace. I find them compelling enough from a dialog generation standpoint – and would like to just influence them directly, rather than try to dialog with them. What would Façade be like if instead of a natural language interface, there was a “palette” (as Chris Crawford would call it) of words and actions available at all times? Something like a context bar at the bottom, where words faded in and out… growing in size with repeated emphasis, shading from bright to dark?
You could even color them so that individual types of phrases were easily categorized – red for angry words, blue for soothing words… green for action, white for “reflection” (repeating something back just to reinforce it – without explicit emotional emphasis). As the state space changed, so would the intensity and colors of the words in this palette – in some ways, it might even act as a kind of “EQ” readout for the dialog at hand. Intense words ahead! Act now! Would this kind of interface make the user “hyper aware” of the state space – to the point that the drama was unsuccessful? Rob and I debated. I ventured that as long as the state space was big enough and rich enough, the player wouldn’t mind. This is certainly true of games like Poker – where the limited number of cards isn’t usually an issue – thanks to the rich interactions they enable once combined with rules and players. Could this simple change (which is really about giving up on Natural Language for the time being - is that so terrible?) provide the player with an intuitive control of the situation without destroying the value or drama of the dialog? Interactivity vs. Narrative Maybe I’m just a games nut and will never be satisfied with story-driven interactions? I certainly don’t buy games like Final Fantasy for exactly this reason. Rob and I debated, again. What is the spectrum of narrative and interactivity really about? An experience like Façade is characterized by story or drama – theatrical dialog that is the result of some interactions between well-implemented simulated actors and a user. On the other hand - a video game that is story-driven often ties the story to specific actions or points in the action. Narrative content is, basically, a reward for particular choices - or (more often) just for getting somewhere. Game-like narratives focus on presenting a well-crafted story (in this case, one with emotionally engaging themes and dialog) - but the interface and feedback (as mentioned above) can be problematic. Narrative-like games focus on generating good activities through refined systems of interaction (eg: strong UI and gameplay mechanics) but often deliver narrative content in a static and boring way. In a nutshell: both are pretty good at their relative extremes, but half-baked in the middle. What does that mean? I'm not sure even I understand it!
One analogy is team sports. You can play in a basketball game, or be a fan in the stands – but being on the bench is a very different experience from either of these things. You can’t really “relax” because you are watching as a potential actor. But you can’t really act as you want – because you’re benched! The sometimes-benched player, basically, is where both of our extremes land us: participating a lot and then sitting through some narrative, tapping our foot… or watching most of the time but then wondering “Is it my turn? Am I on now?” What does it take to reach the middle ground? And would it be compelling... or fun?
Coaching A couple of years ago I worked on a poster for AAAI – as part of the Intelligent Systems demos for the conference. The gist of it was “coaching” - thinking of the AI for games as a coach or facilitator – rather than an opponent or toy. Today, this idea is at the base of all my work. Whether I'm discussing character AI design, or describing mid-level AI tools for generating or balancing gameplay - I'm thinking about coaches.
It's important to be aware of the player - attempting, at least, to match system behavior to player action and perception. Using the sports analogy from above – we see that one way to keep a player engaged regardless of their position on the field or on the bench, is to make them feel like they are "on the team".
Phil Jackson is famous for exactly this type of philosophy. The team is your family, the game is a team responsibility. No player is excused from the dialog – each practice and trip are significant events in a chain that give the team value and meaning. The coach encourages each player to participate. He teaches them that the game isn’t about winning or losing – it’s about being present and contributing your best. Isn't that the message that we need to send to players – regardless of whether we’re trying mostly to tell a story or mostly to build a fun game? Isn't it our job to keep them focused, attentive, and active? Prince of Conclusion Even in a partially polished state, I think that Façade is a great achievement. It’s proof positive that reactive, dramatic characters are within reach. I hope it sparks some creative thinking about how to make them a reality in commercial games. The things I took away from today’s discussion are pretty simple, but important to me and my understanding of the narrative/interaction continuum: o I am ok with mostly watching or mostly doing – but when these two activities are mixed, I am easily confused and frustrated. o It’s hard to coach partial involvement – to create and then maintain a dialog between the player and the environment. It's difficult to send the "you matter" message consistently. The MMORPG people have been thinking about this stuff a lot... using the language of customers and services. The rest of us should think about a lot more. o If games are going to become more narrative, they’re going to need way better writing. The short scene in Façade puts most game writing to shame. o If interactive stories are going to become viable as commercial products, they’re going to need better feedback mechanics. This includes shortening the loop between player input and story/character response as well as providing intuitive and effective interfaces for the interaction at hand. Typing, for me at least, just doesn’t cut it.
o I am excited about Natural Language Generation and not so hot for Natural Language Understanding. I like the idea that characters can talk to each other (and to me) in relatively intelligent and responsive ways – but I don’t need to "dialog" with them to appreciate that. Pushing for Natural Language Processing in games seems like a fine long-term goal… but in the short-term, I’m ok with phrases, palettes, and buttons. As long as the characters respond in compelling and coherent ways, I’m more than ok with artfully-strung-together dialog snippets. Just show me some finesse! Looking it all over, this feels like very little progress – considering all that there is to think about and work on. Perhaps a chat with Andrew will shed some light? Or perhaps, as Rob pointed out - this is as conclusive as a dialog about IF can be? Nothing to do, really, but get back to work!
|
|
