November 28, 2006

A Wiider Audience

by Andrew Stern · , 12:21 am

Raph Koster links to an NPR “Talk of the Nation” segment about widening the market for videogames. Much of the discussion focuses on moving beyond the buttons-and-thumbpad controller interface, e.g. to the gestural interface of the newly-released Wii. (No mention of natural language interfaces for games, though. :-)

Anyhow, it’s a good overview of the issues involved, discussed in layperson’s terms.

23 Responses to “A Wiider Audience”


  1. nick Says:

    As with the Harper’s forum, this is a good step that (relatively) mainstream journalism and media has taken toward consideration of games as expressive, artistic, and serious. No earthshattering insights were unfolded, perhaps, but I bet some humanist dismissers of video games were listening to NPR.

    It is particularly striking that natural language interfaces weren’t mentioned, given that there was a question about the literary dimension of games. Still, I was pleased that Raph pointed toward niche and independent games as he discussed where interesting work was happening. In the same program with the VP from Nintendo of America!

  2. Raph Says:

    Mentioning natural language interfaces didn’t even occur to me, and since reading this, I’ve been pondering why. And I ended up concluding that it was because the vast majority of work done with natural language interfaces hasn’t clicked in a real viral sense… Facade got a lot of attention, but I don’t know that it got more than Food Force or other prominent Serious Games. And no NL-based game has clicked the way that Line Rider or Flow did.

    I wonder what it will take to have that “moment” so to speak…

  3. nick Says:

    As far as making money with NL interfaces, there have been some commercial games: Seaman for Dreamcast and Lifeline for PS2. These, I must admit, haven’t clicked, but they’re out there.

    Besides the huge hit of high-end Façade, there is also so much IF being developed – hundreds of games a year, and in different languages. At least a handful are really good. I think it’s possible to reach new players and readers with some of these, but while the IF community is going strong, there’s been no big breakthrough past that group.

  4. andrew Says:

    ha, I was being a touch facetious when I complained about the lack of a mention of NL interfaces; as much as I hope it will be part of the future of gaming, I agree that it’s probably not clicked enough yet in a mainstream way to merit discussion on Talk of the Nation. My desire to have it discussed there was a bit of wiishful thinking.

  5. Patrick Says:

    Nick, I was actually thinking about IF’s mass market potential just yesterday, and my judgement is that is won’t click because the feedback isn’t consistent enough. I’m a fairly hardcore gamer type, and I don’t particularly enjoy IF because I always feel like I’m fumbling around in the dark. You’ve got very specific tastes, very literary in that sense, so you can parse it, but I’m fairly confident that IF will remian in the same realm as James Joyce novels, only appreciated by a few.

    On that note, I think Storytron has limited appeal for very similar reasons, though it does offer a bit more consistency than IF.

    The Party could have the appeal, but in order to polish it you need to spend lots of money and person-hours, thats the big downside of the architecture.

    Personally, I think there’s a lot of room in the one-button mouse, and moving deeper, the myriad axes of the Wii-mote, and I believe building character AI architechtures with those interfaces in mind is the hot-ticket. Its more likely to get funding anyway.

  6. michael Says:

    Regarding “you need to spend lots of money and person-hours, thats the big downside of the architecture,” there seems to be a general vibe that the Facade architecture requires way too much work to author for. This is because of how damn long it took us to create Facade. And we’d be the first to agree that improving the authoring pipeline is something that needs to happen (and in fact, we’re working on that). However, creating Facade really only took 6 person years (3 years from each of us spread over 5 years). 6 person years to write all the infrastructure (drama manager, autonomous character architecture, natural language processing system, comic-novel-look procedural animation engine), do all the design (with no standard game design tropes to fall back on), do all the authoring. 6 person years is an extremely modest time budget for a modern game. So another reaction to Facade is “wow, what an efficient architecting and authoring process.” As far as comparing authoring in Facade vs. other architectures, I guess we need to see some completed and released interactive stories in other architectures before we can start making comparisons ;)

  7. Ian Bogost Says:

    Michael — about the 6 person years, would it have been possible to scale that work over more bodies? In other words, could you have really done it in one year if there had been four more of you? This is an earnest question and not a snide one… my guess is that it the management/mindshare with other people puts another time cost on development time here, perhaps even a 100% cost (i.e., 6 people, 2 years), etc.

  8. Ian Bogost Says:

    Incidentally, my new column this week at Serious Games Source / Gamasutra also deals with widening audiences and opportunities in the medium, and it also deals with the Wii, but I go in a decidedly different direction from gestural interfaces.

  9. andrew Says:

    Ian, once our new set of content authoring tools are in place (I’m making progress every day!) we’ll be able to parallelize production across multiple team members. Though, not in as severely a “commoditized” manner as typical game production is. Note much of Facade was more uncertain R&D, so it’s apples and oranges when compared to our current and future commercial productions. But even with Facade we could have parallelized a bit, e.g. theoretically giving the animation engine task (~9 months) to someone else working closely with us, or even the ABL compiler itself, once spec’ed out carefully. We did share some of the NL compiler work with a talented student.

    Patrick, regarding IF’s market potential — my take is it’s not the “literariness” per se that’s the issue; there could be a large market for works that deeply and successfully address human relationships, etc. Rather, first, I think IF’s text-only nature relegates it, at best, as niche in today’s media landscape, akin to how comics are and how novels have become compared to TV, movies and the rise of graphical games. (And presumably partly where you take issue with Storytron.) Second, much of IF is puzzle-oriented, and puzzles are arguably a niche form of entertainment, although certain types of puzzles have mass appeal, e.g. crosswords, though few want to play through a whole collection of crosswords. Finally and most important, IMHO the problem that all interactive stories suffer from: they’d be more satisfying with more agency, as well as, I believe, a conversational NL interface (versus the command interface of IF).

    Making an IF that addresses the second and third issues above, but is still text-only, I think would allow it to at least become marketable to a niche audience. Hmm, maybe we should release “literary” (read: text-only) versions of our interactive dramas… ;-) Like the spinoff novels from movies.

    Nothing wrong with being niche, btw (1 2). Don’t think “it ain’t easy being green”, think “I can still get rich if I’m niche”. (Not that we’re in this for the money.) Maybe the “wider market” will really be the expansion of lots of current and new niche markets (I think I remember a Raph post in the past about this). Someone should write a post, “EA is Dead: A Niche-ean Worldview”

  10. Patrick Says:

    Micheal and Andrew: I don’t mean to disparrage your architechture, I’m in the handicapped position of having to draw comparative analysis on engines which have seen limited production histories. I’ll be speaking on the matter in March, so I’m extremely interested to research the logistics of producing content with a variety of drama engines.

    Storytron is probably the closest thing to an IF that isn’t based on puzzles and offers deep agency, and I do think there’s enough of a niche for it to support a number of people with decent supplementary incomes, and few people with full-time middle-class incomes. From what I know about that architechture, a three character drama on par with Facade’s replayability and content could be produced in one or two person-months, though the experience would be radically different from Facade, of course. I’d imagine mix-in behaviors would consitute Roles, Beats would constitute Plot Points with Fate, Joint-behaviors would be reaction scripts.

    You know what, I’m going to produce a small storyworld with Crawford’s engine with the goal of duplicating the context of Facade as an experiment to demonstrate what profound experiential difference there are between the two architechtures. It seems like something I’d need to do to talk on the issue with authority.

    What I’m argueing though, is even if you take out R&D costs in terms of person-months, and you factor in generously light management costs (say managing a small team reduces efficiency by 30%) it still takes longer to author content in ABL than it does in Storytron Script. Writing a joint dialogue behavior simply takes more time than writing a reaction script. I suspect the balancing costs are also greater in the Facade architechture, because every JDB needs parsing parameters that must be fine-tuned through playtesting, whicle Storytron can be balanced using statistical tools because its characters are based on a homogenous data-structure.

    Please correct me if I’m in error here. Also, I’d be very, very interested to hear what your production estimates are for The Party: how many team members, what sort of management structure, how will content be modulated, will balancing issues increase exponentially with the number of unique characters, how many JDBs per character, and how many Beats?

    If you’d prefer to e-mail me rather than detial it in the public domain, that’d be fine also. I’d really appreciate the information.

    RE: Niches, I definetly think that casual games with a low barrier to entry, and potentially, but not nessecarily, lower depth and meaningful thematic content, are leveraging a wider audience and effectivly allowing a whole new ecology of niches to emerge. Lets say 10 million people have begun playing some kind of interactive entertianment in the past year or two. Those ten million all might kinda like Diner Dash, and one out of fifty might buy it, but in that ten million there’s a whole bunch of smaller segments, ranging from the thousands to hundreds of thousands, that would really love something much more specific and niche oriented. For instance, there are probably at least a thousand people who would be devoted fans of gothic romance storyworlds made with Storytron, and there are probably a few hundred thousand people who would love an immersive interactive drama about a party.

    I’m young, so I’m aiming for mass market at this point because I figure my career can ride some feedback loops that way; but I want to get into producing prestige games with more specific audiences as soon as possible. I’m hoping “as soon as possible” could be as early as Q2 ’07. I’m wondering if the Columbine RPG’s finalist status at Slamdance might serve as leverage regarding funding for such titles.

  11. andrew Says:

    Didn’t want this discussion about reaching wider audiences to become another Facade debate, but…

    It’s not clear at all that an atomic piece of content in Storytron (what you’re calling a script) is equivalent to an atomic piece of content in the Facade architecture (there are various types — a joint dialog behavior, a reaction context, a drama management precondition) — in terms of what each operationally represents, what each accomplishes experientially, etc. So I don’t know how useful it is to try to compare them as apples to apples. For me to understand how to compare them would require me to author in Storytron and understand it, which I haven’t done.

    (Btw, a joint dialog behavior can be as little as 15-20 lines of ABL code, which can be authored in as little as, oh, 30 minutes by an experienced author; our authoring tools we expect to speed this up 3x or more.)

    [In] Storytron … a three character drama on par with Facade’s replayability and content could be produced in one or two person-months, though the experience would be radically different from Facade, of course.

    I highly doubt that time estimate, even if the output of Storytron is only in Deikto (Storytron’s current non-animated, diagrammatic-text interface). The cumulative design process alone for Facade — character and backstory design, design of how the drama would be able to change and differ depending on player action, sample dialog writing, etc. — was at least 6 cumulative person-months of work of creating design documents. Then there’s the coding itself, playtesting, design/writing iterations, etc. (I’m leaving out voice recording, any additional animation required, etc. etc.)

    I agree with Michael that we need to see and play with built interactive stories in Storytron or other new architectures before it makes sense to seriously discuss and compare new architectures and approaches; till then it’s a lot of guessing and speculation.

  12. Patrick Says:

    I admit the topic is highly tangential to this post thread, I should have breached the topic in a more appropriate channel, then again comment threads have a way of jumping to tangential topics.

    My experience with SWAT, which is Storytron’s authoring tool, is that you can make a behavioral atom (a reaction script) in about ten minutes if you know what you’re doing, maybe as little as five, if you really know precisely what you’re doing. I can see a near-equivilancy in time estimates being possible, in regards to content-creation. However, am I wrong in suggesting that the interface still adds time to the production process? Deikto’s constraints make it automatic, if less immersive, while any content in your architecture requires rules of semantic inference in order for the content to function, and these rules must be iterated on, a process continued well Facade’s initial release in July ’05. Am I correct that this is an issue endemic to the architecture?

    “I agree with Michael that we need to see and play with built interactive stories in Storytron or other new architectures before it makes sense to seriously discuss and compare new architectures and approaches; till then it’s a lot of guessing and speculation.”

    By “on par with Facade” I mean, “Facade if it were implemented in Storytron”. My intention is to build off the background work you’ve already done for Facade, and see what its like to author and play something highly similar in another architecture. Crawford tells me they’ll have a stable environment ready as soon as late January, so I should have something ready by GDC. Basically, I’ll take all the discourse acts, define them as verbs, and build-up from there. It’ll be an interesting experiment, and should shed some light on the topography of the design space of drama engines in general, and Stortron vs. the Facade architecture in particular.

  13. nick Says:

    Forget Façade implemented in Storytron, I’m waiting to see Shattertown Sky implemented in the Façade engine.

  14. Ian Bogost Says:

    Ok, so back to the Wii ;)

    I don’t know how many folks out in GTxAland have one yet, but I’m still interested in the question I posed obliquely in my recent column on the gadget: why all the faith in gestural interfaces as market-broadeners? Is the rhetoric just an abstraction, with the Wii offering a convenient but fallacious example for an alternative accessible interface? Arguably (and I really do want to hear peoples’ thoughts here) the Wii is pretty damned hard to use. At least Facade assumes (correctly) that people know how to use a keyboard and form English sentences.

  15. nick Says:

    My own sense about the Wii is not that there is anything inherently right, natural, or brilliant about its interface. It’s more the kind of feeling that is captured by an old business consulting rule: “If everything’s broken, change anything.” I’m hoping that any sort of perturbation that requires that designers and developers stop and rethink things when they write new games, or when they port games, has at least a chance of nunchuking us out of the uninteresting rut of mainstream gaming.

  16. Fox Harrell Says:

    It seems to me that with the Wii Nintendo wants to duplicate (on a console) the success they had with the DS and its stylus based interface. It is not clear that the interface is what brought casual gamers to the DS platform though. Casual gaming software had a lot to do with it, along with additional software that bridges casual gaming and other genres such as adventure gaming. Of course the DS has other qualities that the Wii does not (intimate scale, lower price points).

    I wouldn’t completely underestimate the power of the gestural interface though. I haven’t used the Wii’s controllers personally, but I know that button mashing on controllers like the PS2′s is a severe turn-off to many non-gamers. I have also heard several non-gamers mention that the reason they like the DS is because they can use a stylus. At minimum it is a good rhetorical strategy, how many times have you heard the Wii referred to in the media as a console made to appeal to casual gamers? That alone may cause some to pick up the machine. If there is enough DS-like software it just may work.

  17. Ian Bogost Says:

    Nick, this is probably a very healthy perspective. Perhaps public discussions of the Wii will fall into this more measured hopefulness once more people get their hands on one.

  18. andrew Says:

    I haven’t gotten a chance to try it yet, but even if the Wii-mote is hard to use with its initial games, (in theory) the more naturalistic interface, I hope, will eventually be used to create some innovative new games / interactive entertainments, not merely a more natural tennis or swordfight. A more naturalistic interface seems a necessary (but not sufficient) condition for reaching a wider audience for games, and I agree with Fox that the button-heavy controller is a major impediment to that.

    Did the EyeToy expand the market in any notable way? (As an interface, it may not have had enough expressivity to it to make an impact.)

    IMO the most successful new naturalistic interface for interactive entertainment in recent memory was the Furby.

  19. Patrick Says:

    My experience is that it is fairly easy to use, and that it’s visceral appeal combined with good marketing is effictively raising its profile as a console for the mainstream. Just like casual games on the PC which are otherwise disposable and shallow in their dimensions, Wii’s broadening of the audience will make room for lots of niches, and thats a good thing.

    Two fifteen year old girls were watching me play, and were a bit perplexed, with one talking about it on her cell phone. Thats interesting to me, because it suggests that if the right content is made available, and the console by proxy or by direct demographic penentration, then we really can have a strong market for dramatic interaction.

    Speaking of dramatic interaction, 80% of human communication is non-verbal, i.e. gestural (facial expressions and tone of voice also play a big part of that). Imagine a game where the way you held the wii-mote affected the demeanor and tone of the player character, allowing a nuanced performance with socio-dramatic implications. It seems much more bottom up than language processing, and even more intuitive.

    There’s plenty of potential in langauge interfaces, but I don’t think it’ll come into its own until early next decade.

  20. mark Says:

    I think some of the Wii’s buzz is just that input methods have a certain inherent buzz about them. Even single-purpose ones that work reasonably well for the one game they come with get lots of press: the input methods of Duck Hunt, DDR, and Guitar Hero come to mind. A new general-purpose method that works pretty well is something of a leap forward, given that most previous attempts (the PowerGlove, say) have kind of sucked.

  21. josh g. Says:

    I think there’s a significant amount of hyperbole being generated by the fan base around how great the Wii is out of the mindset that the hype is necessary for the Wii to “win” the new round of console wars.

    Not that I’m saying that’s a bad thing, since I probably do the same myself. It’s sort of like party politics, I guess. Your favorite party / console might not be perfect, but you’ll support them as if they were so at least things move in the right direction.

  22. michael Says:

    Patrick wrote:
    while any content in your architecture requires rules of semantic inference in order for the content to function, and these rules must be iterated on, a process continued well Facade’s initial release in July ‘05. Am I correct that this is an issue endemic to the architecture?

    Rules for the semantic parser do not need to be written for each joint-dialog behavior. In fact, we have one global set of semantic parser rules for the whole drama. Theoretically you might want to change some of these rules for each beat (~100 jdbs), but we found that, for Facade, we could handle context-dependency purely in the discourse management layer, by authoring custom proposers. We author proposers for each beat – this is way above the level of granularity of a jdb. Proposers consist of ~50 lines of easily templated code in a rule language. We did not continue to tune the NLP subsystem of Facade after launch.

  23. Patrick Says:

    Thank you, thats very interesting and elucidating for me!

    I accept that as a Christmas gift, its on par with getting a Wii.

Powered by WordPress