Blog Post

How to create an avatar for work-related virtual worlds

The use of virtual worlds is increasing for marketing, training and collaboration, according to a survey that Aliza recently covered. This may mean that there will be more enterprise use for Second Life, Open Simulator, and other online virtual environments. According to another survey, early adopters of virtual worlds are learning to use it for brainstorming and project coordination. This survey also suggests that those who use virtual world technology for recreation are likely to be among the first in their organization to explore the possibilities of using it for business.

As virtual worlds become as prevalent in business as they are in games and socialization, we need to learn how to present ourselves as effectively as possible in this medium. Since first impressions count in virtual teams just as much as they do in face-to-face teams, we need to take our virtual appearance more seriously. In the case of these immersive environments, appearance is based on our avatars.

How do we create avatars that are more than just placeholders? Which avatar features contribute most to the quality and richness of the conversation?

Eye gaze

While 3-D avatars can generate motion and activity, it doesn’t seem to be enough. Just because an avatar looks and moves like a person, it doesn’t mean that the quality of communication mimics face-to-face interactions. Avatars still can’t use subtle body language in the way humans do. But adding even something as mundane as head and eye motion, combined as “gaze”, can make a huge difference.

This was demonstrated in a study conducted in University College London. The research compared an avatar whose gaze was based on the conversation’s turn-taking (informed-gaze) and an avatar whose animations were unrelated to the conversation (random-gaze). They also compared these two types of avatars to other media, such as live video and audio-only conversations.

According to the study, even the inclusion of a relevant, expressive gaze created a significant improvement in the perceived quality of the conversation. When comparing to the participants’ sense of involvement, informed-gaze avatars barely had a significant difference from communicating via video.

What’s even more surprising is that the random-gaze avatar had no significant improvement over pure audio. This implies that a mere 3-D presence doesn’t automatically mean that the communication quality and richness improves. To be truly useful and engaging, avatars must exhibit behaviors that are relevant to the conversation — even if it’s something as simple as eye gaze.

Gesture and manners

Taking a step beyond eye gaze, it appears that other non-verbal manners, and our responses to them, persist in virtual worlds.

In a study from Stanford University, researchers observed how interacting avatars used interpersonal distance and eye gaze in Second Life. Based on their observations, non-verbal norms were still present, even if controlling them meant moving a mouse and keyboard rather than our own eyes and legs. The study, which focused on gender norms, showed that male-male pairs, like their face-to-face counterparts, maintained less eye contact and were further apart than female-female pairs. Also, when the “physical” distance between avatars was decreased, the avatars compensated by avoiding eye contact — just like people would in the offline world.

But it’s not just gender norms that persist in virtual environments. Cultural norms are still exhibited, even when the technology allows us interaction with people from different cultures. An experiment from Augsburg University in Germany showed that cultural speech nuances such as timing and pauses affected a viewer’s preference for an avatar. Their results showed that people preferred avatars that exhibited the nuances from their own culture. Perhaps this means that even if we are meeting other people in a virtual environment, we should still take cultural factors into account when we’re communicating.

How important are these cues? Quoting from a paper on the use of avatars in collaboration (PDF):

The ability or inability to maintain an “appropriate” interpersonal distance or the ability or inability to control one’s viewpoint so other avatars may be seen, among other things, will affect the feeling of being there together. […] Being in an environment where people feel connected is more likely to enhance development of trust and reduction of conflict, which in turn lead to more effective collaboration.

(Source: “Avatars, Text, and Miscommunication: The Impact of Communication Richness on Global Virtual Team Collaboration”, Columbia University)

Looking back at all the research, it’s a lot to take in. Doesn’t it seem cumbersome to have avatars that allow you to control all of these tiny factors? After all, in real life we barely even think about these non-verbal cues.

The answer is simple: it should be automated in our avatars’ programming, just as it is in our own minds and bodies. The more these virtual environments have an intuitive interface, and the more these communication nuances are automated, the more natural it will be for most people to successfully use their virtual world avatars for collaboration.

How do you create and control avatars?

Photo courtesy Flickr user cloud2013

3 Responses to “How to create an avatar for work-related virtual worlds”

  1. This is accurate. We just launched a product for collaborating online with colleagues, the CEO quote was as follows:

    “Web Collaboration allows people to work the way their avatar would if they could get a job on Second Life. If we weren’t afraid of lawsuits, we might have called it “Second Job”. Also, that wouldn’t have made any sense to a lot of people.” – Scott Annan, CEO of Mercury Grove

    We are thinking along the same lines, Glad to see the point of view is valid in many spaces.

  2. Nice post, and I totally agree on the avatar appearance premise. Many people who use virtual worlds socially are (either actually or in their own lack of self esteem minds) not what they or others would call attractive in their physical appearance. They can feel liberated from the tyranny of judgmental people by creating an avatar persona that they can customize to be (what they consider) beautiful — and even sexy — in appearance. Paradoxically, those same people who in their own self defense say “looks are superficial and should not matter” are the ones that work hardest on making CERTAIN that others in a virtual environment find them attractive in order to win their approval.

    The bottom line, of course, is that looks always matter, both physically and digitally. And this is even more so in a virtual environment because it is almost entirely a visual experience (with some sound thrown in for texture).

    The situation regarding eye gaze and proximity, however, is an entirely separate matter. I have read such studies in the past and some of the earlier ones had an initially interesting idea from which they then made laughable conclusions, purely because they were imbuing more personality and function into the digital character than actually existed. Or, they were drawing conclusions from snapshot evidence that was totally misinterpreted. What comes to mind as an example is this now (infamous, and totally misrepresented) snap from 2006 [] (Note: Just in case the picture is removed by the owner, it depicts two newbie male avatars. One is editing his appearance and the other is standing nearby (on a steep slope, by the way) typing a message. Because of their proximity the hands of the typing avatar are intersecting the other at the waist level. This is a totally innocent and common occurrence even now by people who are not yet used to controlling their avatars and are not familiar with their functions. However, the asinine (and obviously anti-SL®) implication from the poster is evident in the caption which reads, “Welcome to Second Life, Now Let’s Grope….”).

    Later studies (as in ones you have mentioned in your post) have taken a more scientific approach and have shown what is already intuitively obvious to regular avatar users; eye contact and proximity can indeed affect the validity of the in-world experience for many, as well as how seriously one person perceives the other via digital persona. It will always be true, however, that many people couldn’t care less how their avatars are perceived by others as they have absolutely no mental or emotional investment in them. To those people the avatar is nothing more than a computer cursor with arms and legs and not good for anything other than as a tool for navigation. Nevertheless, there are still a lot of reasons to be dubious about some aspects of these studies because it’s obvious they are still imbuing the avatar with some kind of magical connection with the human “pilot” if you will. They imply that somehow the user is not “letting” the avatar be animated to respond to a speaker, make eye contact, stand a certain distance away, or otherwise show human body language to enrich the interaction. Nothing could be more wrong.

    Someday we will have holographic avatars that respond to our physical-world movements and expressions, but currently we have a rather crude facsimile to work with. Regardless of how we (including myself) become emotionally and intellectually attached to our digital selves, the fact remains that avatars are merely puppets, and totally dependent on constant attention and manipulation to appear “alive.” Just as when a marionette’s strings are cut and the puppet falls to the ground inert, so the avatar ceases to be a persona when the pilot fails to give it constant attention. What these types of studies never seem to delve into is the WORK that must go into the appearance of body language and to me therein lies the rub, validity-wise. Also, I have no doubt that this is why many find virtual worlds unappealing; they simply don’t want to take the time or make the effort to become master puppeteers (and that’s a totally valid concern — time is precious).

    However, for those readers who ARE interested I have a few comments:

    1) In the default state, without any type of animation overrider or active manipulation, an avatar in Second Life or similar virtual world stands and cycles through a series of minor (rather boring) fidget animations and stares vacantly off into space with a deadpan expression. Not very enthralling. It’s important to know that in Second Life and OpenSim (which is based on SL) the avatar skeleton and facial morph points have been severely reduced, presumably to lighten the resources an avatar needs (which are already considerable anyway), therefore facial expression is yet another entire issue of contention, since one has to “make do” with the limited possibilities.

    2) By adding animations via an overrider that provides more interesting and even entertaining body movements one can personalize the avatar’s presence, but it’s still just hands-off fidgeting, and the eyes are still just vacantly staring in the direction the head is turned with no facial expression.

    3) If one wants to show active interest in anything the avatar must be moved toward that person or thing, the head must be turned in that direction, and the eyes must appear to be focused on the subject. All of this requires practice and in many cases finesse, without which everything from lack of interest to misconstrued actions (as mentioned above) can be perceived.

    4) In the default state the avatar’s head (direction of looking) will follow the mouse cursor on the screen, so in the simplest sense all one has to do to appear interested in someone who is nearby is to turn in that person’s direction and then rest the mouse cursor on their head. Your avatar will then appear to be looking directly at the other avatar’s face. Of course, this effect is easily (and often) thwarted when you move your mouse cursor around on the screen to find things, click on them, or use the interface. Your avatar’s head then swivels all around following the cursor and you appear distracted and not focused (quite the contrary!). By the way, once you are acclimated to it, you can often tell what another person is doing on their screen by watching their avatar’s head movements. :-) A very common result of this lack of practice or concern can be seen when someone’s avatar appears to be constantly looking down at the ground near their own feet, a sure sign that puppeteering is for the moment neglected. The most blatant result of neglect is that the avatar “goes AFK” and appears to fall asleep in-world. Always a big no-no, but a frequent occurence.

    5) To take control over this head/eye movement indicating attention is simple, but requires vigilance and frequent updating (its primary weakness and the cause for much of the misconception in studies). You merely hold down the ALT key on the keyboard then mouse-click on the person’s face (or other object) to focus your avatar’s gaze there. Even if the object/person of attention moves your avatar will follow the focused point until the action is broken by your movement, or by an animation that overrides your intent.

    6) In the case of animations, particularly those placed into an animation overrider that replaces the defaults, the avatar’s head and eyes are taken over by the animations IF the animation creator gives those movements priority. If the animator relinquishes control in his/her animation to the default setting then the mouse-cursor rules above still apply. If not, then your avatar becomes a temporary slave to the current animation and looks away.

    7) And finally (yay, right?) the best way to ensure an appearance of focused attention is to buy or create (and this is much easier than most people think) poses and animations that control one or more avatars so that they are made to look and behave as if they are paying attention (even though the puppeteer may be off looking at Twitter and only listening to the discussion). An example would be a dual pose on which each avatar sits which gives the illusion of them both standing in front of each other gesticulating and looking each other in the eye, or a set of classroom animations where all the attendees are apparently paying rapt attention to the speaker in the front of the room.

    As you conclude in your post, to make communication and collaboration in-world more effective and comfortable spontaneous activity and personal space issues would need to be automatic. However, that is far from the case presently, and I think will be so for quite some time. The more important question, it seems to me, is how important is this interactivity to each person and HOW MUCH WORK are they willing to put into the implementation of same?

    By the way, if we ever meet in Second Life, I’ll expect you to look at me when I’m talking to you. ;-)

    • Thanks for your insightful comment – I think it’s even longer than my article! But you’ve raised a lot of interesting points.

      “It will always be true, however, that many people couldn’t care less how their avatars are perceived by others as they have absolutely no mental or emotional investment in them.”

      I agree, especially if the use of the avatar is recreational and/or infrequent. The assumption I’m making in this article is that the reader is interested in using an avatar for business. If that’s the case, then they should have at least some mental investment, since it’s a requirement for the work that they do.

      “They imply that somehow the user is not “letting” the avatar be animated to respond to a speaker, make eye contact, stand a certain distance away, or otherwise show human body language to enrich the interaction.”

      Not really, since in most of the studies there is no human controller and the avatar’s liveliness is preprogrammed (of course this isn’t true of the studies based in existing virtual world activities rather than a limited study in the lab). This has led to some interesting results, such as the inferred eye-gaze avatars being seen as more “real” even if no one was controlling them. (Here’s another one in a similar vein, but it’s kind of old:

      “What these types of studies never seem to delve into is the WORK that must go into the appearance of body language and to me therein lies the rub, validity-wise. Also, I have no doubt that this is why many find virtual worlds unappealing; they simply don’t want to take the time or make the effort to become master puppeteers (and that’s a totally valid concern — time is precious).”

      This is true, and it’s not just time that is precious – so is attention. It’s hard to focus on the “realism” of your avatar if you’re in the middle of a meeting or collaboration study. Also, to address your point about how much work it takes to implement this “realism” in one’s avatar, there is this other study that I failed to include because of length constraints, but you might be interested in it:

      Basically, it shows how the addition of text-based chat, accessories or other visual cues in place of things like raising one’s hand, voting, etc. There’s a lengthy part about seating confusion, when the users were still figuring out how to get their avatars seated correctly.

      Like you, I do want to see more reliable studies on avatar use, since I couldn’t find many that explored the learning curve or the “breaking in” it takes to get used to handling an avatar.