Graphical Conversations



Using ComicChat is very easy: since its based upon IRC, it leverages off of the already learned chat interface. But its ease of use is also due to ComicChat's very limited expressive abilities. Each person is represented by a well drawn comic character, and the handful of expressive forms that the system supplies to the user are useful, but they are very discrete. Part of ComicChat's charm is that it creates an aesthetically pleasing environment in which to interact with others, which is in marked difference to other avatar based chat. Much of the work of laying out the conversational environment is done for you, which aids in maintaining a visually pleasing environment. None of the other avatar systems do this much work to place characters and generate expressive stances for them.

The emotion wheel is a particularly novel method of controlling the emotive representation of the avatar. It's functionality was immediately apparent, and its use wasn't even necessary most of the time, since the system did so much work trying to pose and emote characters for the user (which it seemed fairly accurate at doing--but it only had a very discrete and small set of emotes to work with).

ComicChat also has a trait that is shared with its text based predecessor, but not with any other avatar chat system: the history of the interaction is the actual conversation in the same interface as the conversation. Even Chat Cirles forces you to use a different interface to see history information. But ComicChat does one better than IRC: it makes this history and placement immediately visually apparent. Since each frame is visually distint and unique, the process of visual search works quickly to find where you might have left off in the conversation.

There are of course some problems: time in ComicChat is compressed; while turntaking is preserved, the time between each turn isn't. For example, there can be a 5 second or a 20 minute pause between two panels, and ComicChat would portray them in the same way. Emotive content is extremely limited; There is no in-between between chuckle and hysterical laughter, and though its not clear IRC or any other text based chat has conventions for this distinction either, the ability of IRC to emote any given text provides a great deal of flexibilty. In ComicChat, the tight coupling of emoting and visual expression of emotional content makes either saying or thinking that you are "laughing" or "chuckling" very displeasing and strange.

Identity, while more obvious than in a text chat, is also rather generic. Since the avatars in ComicChat are coherent and organic, they are easy to identify and express their limited emotional content well. They also serve as an easy way to keep track of participants at a glance, since each character is distinct, unlike IRC usernames, which all look visually similar. But as the first few panels above show, it is easy to have two people with the same avatar, at which point it is impossible to tell the two people apart visually, since the avatars are so generic. Also, panels show no sense of conversation threading. Every conversation flows together in this system, unlike Chat Circles, which does a very good job at pulling apart individual conversations within a larger space.


Interacting using OnLive was a bit more difficult than with ComicChat. First, I had to choose and modify my avatar, then, once in the environment, I had to figure out where people were and how to navigate myself to a conversation (which was especially difficult since there were only a handful of people on the system. Navigatin in this environment is awful. I floated very slowly around the room, and didn't go up over objects automatically. Everything is left to the user in terms of interacting in this system, unlike ComicChat which provides a great deal of help and default behavior.

The Avatars do have some emote ability, in the form of changing expressions on the faces of the disembodied heads (which were very eerie to interact with). However, the expressions are limited, and to provide anything other than a blank stare, you have to constantly be changing your face. Interaction takes a great deal of effort in this environment.

The novel part of OnLive is that it is a voice interface, so using the keyboard for emoting is a feasible requirement. However, talking to strangers is a much more personal experience than chatting (at least initially). Creating a false identity (especially male posing as female) becomes more difficult since your personality is tied (perhaps) more closely with verbal than with textual interaction. Maintaining a conversation with OnLive is more difficult than in text chats since the space between speech is less tolerable than between text posts.

However, with all of its faults, since the interface for interaction is multimodal, OnLive creates a richer interactive environment with greater ambiance (use of background music) that is more believable (forgetting it's awful visual interface) than text based avatar systems.

Designing a Better Graphical Environment

One question to ask is "why have 3D chat worlds not taken off?" Perhaps the most obvious and general reason is because they're not compelling in long term. While it may be fun to build some piece of architecture, or novel to shop virtually, or perhaps freakishly compelling to talk to a large eyeball within a mouth, the novelty of these systems wears off quickly. IRC has staying power because it gives users something they can't get anywhere else: the ability to talk in real time with other people throughout the world. And while you can do this on virtually any avatar based chat system, the interfaces, both navigational and conversational require too much work both to use and to believe. This is where the simplicity of IRC shines. In the end, it is for the social interaction that people use chat systems (avatar or otherwise).

The key, I expect, is not to duplicate what a person can have in RL nor what another system offers them OL. ChatCircles accomplishes this to a certain extent. Its simple interface allows people to just use it without really thinking about how to move about or what they look like. It falls short, however, in expressing emotional content. Where a visual display can really pull ahead of text based chat is both in organizational capabilities, leveraging off of well founded (and in some cases, researched) visual abilities of humans in visual search and sort, and in an expressive interface for emoting (though this must be implemented carefully) .

I found that the most compelling 3D system is the first person shooter, such as Quake or Unreal. Part of the compellingness of such games is their goal oriented nature. But they are also visually stunning, going far beyond any 3D chat I have seen. Though they are resource intensive programs, we have to expect that realistic (or somewhat so) 3D interfaces are going to require great computing power. They of course have no emote-ability, but the fluidity of navigation is light-years ahead of such systems like OnLive. Any 3D chat environment must be as fluid if it is going to be usable on a regular basis.

2D environments are technically easier to create, but they have problems of their own. In a place like thePalace or OnChat, the background graphics mean nothing beacuse they don't effect the conversation. ComicChat backgrounds are at least reactive to what is said, but people still cannot have any interaction with them. For a 2D space to be believeable side view avatars, it needs to provide some sort of structure to allow people to place themselves understandably in the space. Having people free-floating above a set of houses might be amusing for a couple of minutes, but it adds nothing to the conversational environment, and serves as what Tufte calls "Chart Junk". Chat Cirles skirts this issue because it is a more top down view of the conversation, but it too could use a more structured spaces such as private corners to talk privately with a couple of people or podia on which to speak to a larger audience, instead of the generic and visually bland black background.

Emoting facility and fluidity is perhaps the most important feature of a social chat environment. IRC provides a basic but surprisingly capable way to emote. No graphical chat system provides such an open tool. ComicChat's emotion wheel is a step in the right direction; its ease of use and fluid control of (limited) emotional stances is welcome. However it is severely limiting in the emotes it can portray, and the positing of a comic form cuts off the ability to use text to be expressive. OnChat's bodyless heads is perhaps the most disconcerting interface I've ever encountered. Adding to the oddity and absurdity of heads floating about in space is their complete lack of emotional expression.

The problem, of course, is that as soon as you move to a graphical system with avatars that represent real people, you dredge up user's many varied and deep seated expectations about how they will react visually. Unless you are going to provide satisfying, understandable, and expected emote capabilities--and the lack of any emotes in such a system actually means something when dealing with avatars--then the chat system will seem unreal and frustrating to its users. Again, ComicChat is the only system that accomplishes this on some level, but its limitations are a frustration in themselves. The situation becomes even more important in 3D environments because they are the ones that are infused with the most expectations about human behavior. I must honestly say that I am not even close to thinking of a solution to this problem. Gesture is needed, so perhaps the next step may be to use a system such as Flock-of-Birds to allow people's real world gestures to translate to the virtual world. Such a system has an immediate problem if people are going to be typing, since they can't be gesturing at the same time. A facial expression recognizer might be useful in this situation. The key in this case it to separate control of emoting from cortrol of chat. If chat is textual, emoting must either be very easy and fast with the keyboard, or must be entered through a non textual (non keyboard) medium. If chat is voice, then emoting can be transferred to the keyboard (or some other manual device). In a sence, people will have to become puppeteers in a very real sense for 3D chat environments to work.

For 2D environments, expressive capabilities must be increased. In an abstract system such as Chat Circles, there must be some visual way to express emotes. For avatar systems, the actual avatars have to have enough varied, realistic expressions in order to be able to allow users to accurately express themselves.


PoserChat is an interface for an Avatar based chat system that gives users generalized control over posing and emoting through an easy to use interface.

Click for a brief description and mock-up screens.

page last updated 3/27/2001