 |
Federated Presentation Spaces
Computer mediated presentation systems can offer both some substantial opportunities as well as present significant challenges relative to traditional physical presentation sessions. In this work, we will explore a number of ways to enrich the experience both of presenters and audience members in computer mediated presentations using a virtual world interface.
In particular, there are a number of problems with existing mediated presentation systems that we would like to address. Our most direct experience is with virtual worlds, but most of these problems are shared with mediated presentation systems in general:
- Static presenters: Presenters in virtual worlds tend to leave their avatar motionless because of the cognitive load of both presenting and managing an avatar at the same time.
- Dynamic audiences: In a mediated environment, much of the motion of avatars in the audience is not socially significant (in terms of engagement or signalling) and can be very distracting for the presenter.
- Communication channels: Audience-audience and audience-presenter communication is sometimes conflated, sometimes impossible, but rarely is the scope and timing of communication easily manageable by participants.
These challenges also suggest a set of main design questions that — if answered — might address the main drawbacks of mediated presentations. These questions are:
- How can we represent the audience (and its behavior) to the presenter in a way that is natural, minimally distracting, and meaningful?
- How should the audience perceive the presenter's performance of the presentation?
- What kinds of activities can the audience do during a presentation that enrich their own experience and the experience of other audience members?
It is also essential to us that a mediated presentation system be able to handle the same numbers of audience members as a small auditorium; 100-200 audience members should not an unreasonable target and systems we build should be able to scale to that number both as a user experience and as technical limit.
The vision for this project is to provide a set of tools that the audience can use for meaningful and relevant interaction with other audience members during the presentation. We can instrument audience use of these tools to provide more accurate representations of audience engagement to the presenter to help the presenter understand how the audience is receiving their presentation. We can also use a presenter's interaction with information about the audience as a signal to the audience about the presenter's engagement. This will hopefully address the primary problems we identified with mediated presentations. With a good set of audience tools, the dynamism of audiences will be focused in a way that can be made meaningful to the presenter; presenters can more richly signal their engagement with audience information, and federation will better organize conversation among audience members.
Audience Tools
In physical spaces, audience interaction is typically seen as detracting from the presentation itself; whispering to someone next to you is viewed as a distraction and disrespectful to the presenter. In a mediated space, though, the social meaning in doing things other than listening to the presentation can be re-imagined. Participation in audience tools need not be disrespectful or distracting. In fact, use of these kinds of tools can enrich the audience experience as well as provide cues to the presenter about audience engagement.
I've included below a list of various kinds of audience tools ideas that I've generated with the help of Jordan and Nicole. I'll probably focus in on one or two of these in the end, so it's useful to think about them in terms of:
- How much would you want something like this as an audience member?
- Would use of this kind of tool provide useful signals to the presenter about the audience's engagement?
- Is it feasible to use a tool like this with potentially 100 other audience members?
Audio Grabber
In physical presentations where the audience is using Twitter as a kind of backchannel, people often post favorite snippets from the speaker. (eg http://twitter.com/krogull/status/2134476551, http://twitter.com/m2sE/statuses/2132203869, http://twitter.com/cstubbs/status/2135945616) This desire to share interesting bits of an experience with others is a potentially powerful way to encourage audience participation. While listening to a presentation, audience members could mark chunks of audio, both for their own use (for instance, to remember the name of a project they're interested in learning about later), but the act of saving audio can be aggregated across audience members to build a kind of heat-map of parts that drew particular interest from the audience. In this way, the system would both provide audience members with their own personal favorite moments of a talk, but also provide people coming to the audio transcript later a kind of roadmap to help them replay the event.
The simplest version of this would just be a one-button book-mark style interface, in which pressing the "Save Audio Snippet" would just mark the audio as having an interesting point that you want to come back to later. A more elaborate system might allow you to specify a duration, and enable a little bit of light audio-editing on the fly to specify a start and end to a clip. This system could also grow to be more delicious-like, letting people annotate periods of the talk with tags and descriptions.
How the input UI looks depends a lot on how elaborate we get, but a quick mockup of the marked-up audio output is shown below.
If we had durations of clips and some words about them, we could do a really neat sort of linear tag cloud showing what words were used to describe that section of the talk over time, eg:
This shares some elements of the the Audio Notebook(http://www.media.mit.edu/speech/people/lisa/anb.html) notebook project.
Annotation
Audience members often see connections between the content of the presentation and other related ideas that aren't covered by the presentation itself. A presentation annotation system could support the sharing of content and questions related to the presentation itself. With this system, audience members could post relevant URLs or images.
Mockup of such a system is below. Hitting the + button would pull up a dialog box where you can copy in a URL or upload an image and specify a title for your post. It might also make sense to have some sort of up-voting mechanism on this, as well as showing most-clicked links.
Audience-Aware PDF Viewer
Augment the PDF viewer so it exposes how many people are looking at a particular slide. This is a useful signal to the presenter about what parts of a presentation are holding people's interest, or perhaps would show when people had disconnected from the main presentation and not re-engaged. Mockup below.
Hand-Raising
There are lots of different ways we could handle question asking and hand raising. If we want to support something analogous to the way it works face to face while still handling lots of people, we could build a queue-standing system where avatars in line from your local world are fully rendered, while avatars from other worlds are shown as shadows in the line. Mockup below, local users in green.
It might also be helpful to have some sort of system for submitting questions in advance, or even just adding a little bit of metadata about your question (eg on topic, new topic, comment) to help the person picking who should ask the next question.
Note Taking / Question Asking
Instead of users grabbing bits of audio, perhaps they could instead work together to take notes about the presentation? As with the audio grabber, this might have a similar public/private model: users maintain their own notes files for their personal use, but notes files can be aggregated across the entire audience of note-takers to show what parts of the presentation generated active note-taking. There might also be some fun analysis to do that pulls out popular words or phrases from the group note-taking. The trick here is to make note-taking in the system better than just doing it outside of Wonderland. The killer app here might be association with the core audio, so it's easy to move back and forth between your notes and the original audio. Or perhaps personal notes can be interleaved or cross-referenced with group annotations or chat?
This may well make sense as a component of a general annotation system. The display of audience-generated text should probably be different than images/urls, but there might be some way to unify them. It might also obviate the tag-based version of the Audio Grabber concept. We could build pretty much the same visualization from group notes.
Clapping
This is a nice, recognizable social signal and it would be easy to import it here. I would be particularly pleased if we could make it analog enough that you could differentiate between polite clapping and intense fan clapping. Maybe something having to do with how long you hold the clap button down for and how often you press it?
Profile System
People are always interested in learning more about other people attending events with them. In conference settings, nametags with names and affiliations help provide some of that sense, but even with nametags it's hard to browse an audience and get both a high and low level view of who's in attendance. To support this, we could build a really simple profile system where people can enter metadata about themselves including affiliation, title, discipline, geographic location, etc. I'm not particularly interested in going down the networking support tools path, but I do think people are interested in being able to dig down and learn more about someone who asked an interesting question or posted something interesting to one of the audience tools. It might make sense to plug into some other tool (LinkedIn?? Twitter? Facebook?), but I would be inclined to just keep it simple and mostly free-form. People can link to those tools in their profile if they want.
Chat Circles
Trying to have an audio conversation with another audience member while a presenter is also talking might prove to be too difficult, even if you can avoid the social stigma of talking over someone else. Text conversations are easier to manage, but right now they are sent to the entire world. We could support distinct text conversations in much the same way as audio; spatializing distant text and providing a cone-of-silence-like mechanic for chat. Puzzle Pirates has a neat implementation of this that's worth a look: http://www.puzzlepirates.com/Walking.xhtml.
We might also want a way for people to have a shared chat channel across worlds. There might be something interesting in trying to do some sort of threaded chat that gets baked into the presentation at the end, so people who weren't there can see the presentation and the conversation around the presentation at the same time?
Presenter Tools
The main challenge with presenter tools is to provide an experience that is not too distracting, but provides enoug feedback from the audience that presenters rely on to judge the audience's involvement with the presentation. With that in mind, we don't want to add too much to the presenter's UI and view beyond what is normally there — current slide, next slide, and perhaps a timer. The presenter tools described here focus on helping a presenter gauge the engagement and makeup of their audience in different ways.
These information streams should probably be made available to the audience as well. In a physical meeting, audience members can judge each other's engagement, and there's no reason to stop that from happening in this presentation space. It's also important that audiences know how they are represented to the presenter so they can appropriately manage their own representation.
Visualizing The Audience
There are two main approaches we're thinking about for portraying information about the audience to the presenter. The first is a visualization of the entire audience. Inspired by Wendy Kellogg's work on Social Proxies, we can imagine a space in which people present in the meeting (across all the connected worlds) are represented in some abstract and low-resolution way. These proxies for audience members can have their properties changed to represent some feature of the person they represent. They can also be moved around the space. So to steal shamelessly from Kellogg's lecture model, some quick mockups below. In the first two, imagine that we map engagement/participation onto the distance of a dot (representing a single person) from the center of the room. In this sketch, there are lots of people who are not engaged at the back and a handful of engaged participants at the front.
We might also want to change the color to show which federated server someone is coming from, which might bring out some trends in the relative participation of different sites.
Finally, we might be curious about the geographic distribution of users. In this sketch, the dots are distributed and sized based on how many people from that site (per their color) are in particular parts of the country.
These site-specific visualizations depend on each world participating in the federated presentation having some meaning to it. There are some situations where this makes quite a bit of sense, for instance different divisions of a company might each have their own Wonderland instance that they use to watch presentations from. The model might not make sense for a presentation open to the general public, though — it's unlikely that which server people pick to watch from has any particular meaning. we can also imagine organizing the audience based on information they entered into a profile system.
Although avatars are the primary method of representing people in a virtual world, it may not be technically feasible to continue to use full size and resolution avatars to represent a space with a large number of avatars. In these sketches I've been using circles for simplicity, but we might use something like the orbs for users calling into Wonderland, or flat screenshots of people's avatars.
Sonifying the Audience
A high detail visual representation is good for getting a macroscopic view of activity as well as for drilling down to see the microscopic behavior of specific people. For instance if you were trying to choose which person to answer a question from, you might be inclined to pick someone who was engaged for the whole talk over someone who was less involved, or you might try to balance questions between different servers. Presenters might not have that kind of attention available, though, and a more ambient audio representation could be more effective at representing infrequent events or a general sense of the audience's overall involvement.
Audio is a natural channel for engagement feedback because it already plays a huge role in back-channel communication in both one on one conversations and presentations. We could easily map audience participation in audience tools onto low-volume sounds that the presenter can hear. For instance, lots of chatting between audience members could be represented as a low hum; annotations and questions could be bells, and audience members joining or leaving could be doors opening or closing. This also helps with the disconcerting experience of presenting to a totally silent audio channel. We're used to hearing something back from people we're talking too, and this provides some meaningful audio without people saying anything.
Although audio seems particularly promising, it's certainly possible to design visually ambient displays that map the same participation data onto the presenters screen in some way. I think we'll need to play around with different approaches to see what works if we go down this path.
Presenter Visibility
As with audience tools, a presenter's use of presenter tools can be made visible to the audience as a way of demonstrating the presenter's engagement. In much the same way that an active and engaged presenter makes eye contact with audience members and stays aware of an audience's interests and questions, an active presenter's use of visualizations of the audience can be reflected back at the audience itself. One easy way to support this would be to provide low-resolution engagement information (eg don't show individual users, just show an aggregate engagement metric for each connected world) and display individual users only if the presenter mouses-over that section. Audience members could then see when the presenter was looking at them directly. This would be particularly natural during question asking. The presenter might want to see more about who's asking a question (or to choose which person to call on), and so would naturally drill down on that display, which would double as acknowledgement from the presenter that they were paying attention to the question asker.
|