ScienceOrganizer:

Dan Berrios (NASA Ames): The experience of the crew and RST using ScienceOrganizer should have improved over the two weeks of this rotation. We experienced problems with Tomcat, the program that supports the web execution of ScienceOrganizer, early on. As a result, some of the data collected in the first EVAs did not get stored and linked into the system. We corrected some of the programming of the ScienceOrganizer Communications Agent, and Tomcat appeared to be stabilized thereafter; no further loss of data due to this problem occurred.

For future rotations, we'd like to create a truly collaborative system using ScienceOrganizer, one in which the RST can create, modify, and annotate items in the same installation of ScienceOrganizer that the crew uses. This would allow the RST to store and share with the crew their comments regarding the EVA collected science data as the data is collected. We'd also like to explore the development of various notification systems using email or semantic web broadcasting, so that items of interest can be pushed to RST members if they so desire, rather than requiring them to pull the data.

We'd also like to rework the science data models that ScienceOrganizer and Brahms currently share. It is not clear that one model is best for both human and agent users. We could also participate in the development of the speech interface, specifically as it is used to link (associate) items in Brahms. There may be benefit for the RST to further specifying what type of associations two items have. For example, voice annotations can be recorded at a location, requested at a location, or describe a location. Soliciting which of these linkages is the intended association and making that specific link in ScienceOrganizer could clarify presentations of the science data and help the RST locate data more specifically.

Meeting Replay Tool:

Danius Michaelides & Kevin Page (Univ. Southampton, UK): On the final day of the rotation, Maarten managed to video the morning EVA planning meeting in the hab - a change to the usual evening analysis meeting. As we've been learning, it's useful for the RST to have as much information as they can about the crew plans, to give them some context for the data as it arrives from the EVA. As such, this seemed like another good opportunity to use a Meeting Replay.

The Meeting Replay web interface: A digital video (top right) is made of the Mars crew discussion, while Compendium maps (main view) are projected in the Hab as they are edited on the fly to structure and capture the discussion. The video files of the crew and the Compendium screen are then processed and uploaded by the Crew Uplink Lead (Maarten Sierhuis) to a NASA streaming video server. These are pointed to from the Meeting Replay interface created by Univ. of Southampton (Danius and Kevin), who synchronise the videos, annotate the speakers, and integrate an RDF export from Compendium. Within a few hours, the Earth scientists can replay the crew's meeting individually or together over the internet, in order to feed back on the plans and data analysis. The meeting can be navigated by any node in the Compendium map, or by speaker (see the timelines in bottom frame).

Maarten was burning the midnight oil again, and it was late morning in the UK by the time the video was uploaded to the streaming server (not necessarily a bad thing on a Saturday!). So it was a real squeeze to get the replay generated before the RST met - but we managed to rush something out. Also, because of the different structure of the meeting, the Compendium export didn't provide as useful a navigation aid as we've become used to. Still, this was compensated by the video itself, which was somewhat more entertaining than usual... perhaps life in the desert had finally got to the crew!

In the last fortnight we've gained valuable experience using our tools to support distributed scientists - the RST - which is what the CoAKTinG project is all about. We've seen how the replay enables the Mars crew to capture their thoughts and plans, and that it's genuinely useful for the RST to have a detailed replay they can use in conjunction with Compendium. We've learnt from experience how tight the timescales are on a mission like this - hopefully the collaboration tools we're developing will continue to help both the crew, and the remote scientists, make best use of the time and information available to them.

BuddySpace:

Simon Buckingham Shum & Marc Eisenstadt (Open Univ., UK): BuddySpace was released mid-week 2, and as the screen from the final RST teleconference on 8th May shows, most had it installed, with their presence displayed live and active, and a groupchat instant messaging room ran in parallel with the WebEx meeting and telecon . If someone loses their internet access, this is immediately obvious if they disappear from BuddySpace.

BuddySpace presence and instant messaging support in an RST meeting: The Earth-based scientists, distributed in time and space, need to coordinate and communicate. BuddySpace superimposes 'presence' status onto useful visualizations, such as this geographical map. Scientists can launch 2-way or whole-group instant messaging sessions by clicking on the interactive map. Instant messaging provided a valuable 'back-channel' for participants to exchange private messages without disrupting the main discussion on the teleconference, while the shared groupchat room allowed all members to post URLs and nodes to the Compendium map.

The RST facilitators Al and Simon showed how the groupchat room can be used as a simple 'group submission interface' to Compendium. Any participant who wants to add a node to the map can type in its title into BuddySpace, which the Compendium mapper can then drop straight into the map.

It is premature to draw strong conclusions, but now that it's installed on everyone's desktops, the RST can continue to test BuddySpace in the coming months as it considers the implications of the last two weeks. The map can be updated as membership/locations change (e.g. now that Maarten is formally no longer out of RST communication as he was in the Hab, we will add him), which will update each team member's map automatically.

Compendium:

Simon Buckingham Shum, Al Selvin & Michelle Bachler (Open Univ., UK): On Saturday morning, I (Simon) went through what had by now become the standard process of generating a Compendium web portal map (see the May 4th RST-CommSys report), ready for the US RST members when they woke a few hours later in the early hours (the RST telecon is set at 5am California time). While the RST facilitator role had evolved to include this, Al and I reflected that this is in fact a process that could be at least semi-automated, if the right connections were made between the NASA WebEx meeting booking system, and ScienceOrganizer which holds all the EVA data, and stores all the project emails (key ones could for instance be tagged in the subject line as critical for the RST meeting, for inclusion in the portal).

Discussion in this final RST meeting focused largely on reflecting on the preceding two weeks, the time-pressures on Crew and RST (see Shannons' final RST Report), and in this context, how the tools could be improved. A critical theme to emerge was the way in which science data is organized, with some very interesting ideas emerging on the best ways to navigate data, and of particular interest to us, how the spatial/semantic hypertext interface offered by Compendium could support the work practices of the RST members when they are being asked to make sense of huge amounts of data in a short timeframe.

We mocked up a simple example of an improvement, taking one of the Brahms-generated science data maps which embeds each voice annotation and image in a separate map, and moving these into a single map so that they are all shown at a glance, and inspectable with a double-click, rather than having to drill down into separate maps and back up again. Simple things like displaying the relevant images with a voice annotation (and vice-versa) reflected a 'mantra' that emerged from several RST discussions: "join up the ears and the eyes!" (of the astronauts as they capture their data, so that the RST can relive their experience).

Improving science data maps: The science data maps automatically generated by Brahms agents embed each voice annotation and image in a separate map. In their final session, the RST prototyped an improvement, moving all media into the top level map so that they are all shown at a glance, and inspectable with a double-click

To summarise: This MDRS rotation has seen Compendium used in many different ways by a diverse user community. Maarten was the Compendium expert embedded in the Hab, using it as a formal tool to specify the EVA plan in a way which can be read by both the Brahms software agents, and visually by the Crew. He also used it to capture open-ended discussions about the data gathered and what the next day's exploration should be. Abby and Brent have used Compendium in creative ways as a canvas to lay out sample photos linked to analyses and questions, into which the RST has then inserted its own replies and queries.

Meanwhile, within the RST, Al and Simon as facilitators have mapped RST telecon/WebEx discussions in real time, providing a shared display of the emerging issues and ideas for all to see. Compendium portal maps provided a way to filter information to enable the RST to get to prepare for meetings, and web exports of the databases from Maarten allowed them to get to grips with the material without having to locate, download and import sizeable files. Stacy and Brett used Compendium to create their own maps to support their analyses, and in particular, to lay out the evolving SEMS exploration methodology systematically.

It was deeply rewarding to see Compendium operating as an integrated modelling, sensemaking and dialogue environment, interoperable with other key technologies in the worksystem (ScienceOrganizer; Brahms; the Web) , as we have always envisaged it.

Reflections on the RST Facilitator role:

The introduction of new technologies to support the RST also led to the creation of the RST Facilitator role, played by Al and Simon. Here's our first cut at reflecting on what playing this role actually entailed, and how it might develop.

We came into this field test with a number of research questions, including: Can this role and these tools add value in an environment already well supported by other sorts of knowledge tools? If so, what kinds of value? How to characterize the value? What opportunities for further development of tools, methods, and human skills will enhance the value or remove impediments?

A technical objective was to engage the tools and methods in their current state of development in a field test, from an action research perspective - our experience is that doing so always surfaces issues and ideas that can be built on.

The primary tool we used was Compendium, (a) since this was most familiar to us, (b) it was being used with the Hab Crew by Maarten, and (c) it has been designed from the start as a sensemaking/analysis environment, which we hoped would complement ScienceOrganizer's powerful archiving and search facilities.

RST members and facilitators operated in a 'loosely coupled' mode. We worked as a truly virtual US-UK science team, from our separate offices and homes, having never met in person, picking up our tools and data from diverse emails and web links. The facilitators essentially performed a 'knowledge crafting' role, involving 3 kinds of visual mapping:

• creating maps before RST meetings (e.g. portal maps which filter and organise material)
• mapping ideas and discussion during the meeting
• refining the discussion maps after the meeting for publication and dissemination to the other parties involved, as well as archiving

Each of these mapping modes has its own set of considerations and skills involved (which we have written about in papers on the CompendiumInstitute.org website).

Another part to the role is in reflecting on how knowledge management is going in the RST, and at times raising process issues for discussion, such as what's working well in the present moment, what could be improved now or for the next session, but also what could be improved for future RSTs (software or method enhancements, for example)

We're conceiving the facilitator's role as being one of adding shape and coherence. Glueing together important pieces, making significant connections. The facilitator must work in several dimensions at once: listening to what's being said in a meeting, prompting for clarity and cohesion, staying in the background much of the time (the RST are the domain experts). Paying attention to the representations is central: what is currently displayed, following the RST's discussion to navigate the hypertext and display the relevant material as they speak, thinking both about the raw science data, its preliminary shapings (ie. the maps that Brahms agents generate), and the previous sensemaking done by the crew. The facilitator must be expert with the tools, whatever they are, whether an image browser, PowerPoint, the web, or Compendium. In the latter, one is working on coherent labeling of nodes, metadata evolution and tagging, node typing, node placement and linking, so as both to provide the RST with an unfolding canvas of their current discussion as well as shaping the representation for easiest assimilation by the crew.

Could all this be done without the software tools? Yes, but at the cost of each RST member having to do their own piecing together, with no way to integrate them. Could it be done without a facilitator? Again, yes to a degree, but what we believe we saw, borne out by the RST members' comments on numerous occasions, is that significant value is added when someone is dedicated to crafting effective collaborative 'knowledge representations' (in the broadest sense). In the highly compressed time-scope of a mission, the RST facilitator seeks to minimise mission time wasted on low-value-added activities (finding bits of info, re-creating the sense that others have made of the bits) and maximise the scope for domain experts to make their unique contribution: coming to grips with what's most interesting and significant in the huge streams of data arriving on their desktops, and adding value to the Crew's deliberations.

So the role of the RST Facilitator could be said to be "significance highlighter" or "significance enhancer" - filtering, highlighting, arranging, crafting so as to provide the best launching pad for RST analysis, as well as subsequent use and uptake of the materials by other teams.

Conclusion:

The RST-CommSys Team was a very fruitful human and technical collaboration between the ScienceOrganizer and Mobile Agents teams at NASA Ames, and the UK CoAKTinG Project team. Open Univ. and Univ. Southampton are indebted to Maarten Sierhuis and Bill Clancey at NASA Ames for inviting them to work on the challenge of evolving effective Mars/Earth human-agent worksystems. What better testbed could we have hoped for! We look forward to continuing this collaboration as we work on the many ideas and data generated.

And since we've never met many of the Mobile Agents team except in cyberspace, here's the OU team bidding you adieu, for the moment!...

The OU team: Compendium: Simon Buckingham Shum, Michelle Bachler & Al Selvin (inset, based in New York). BuddySpace: Jiri Komzak & Marc Eisenstadt