Overview

I propose to perform a human factors assay to determine the feasibility of developing an interactive crew advisory system suitable for deploying in hab, on an ATV or on a personal digital assistant. During the project, I will monitor crew activities, compare information needs and available resources, and assess what kinds of assistance would be compatible with crew workflow. The deliverables of this project will be an initial human factors assessment, development of standardized observation forms to continue the assessment with subsequent crews, and an initial requirements document that can guide further research into off-site development of the technology platform for the advisory tool.

Motivation and Approach

Operating procedures of a space habitat are complex, and it is impossible to plan for every contingency, so crews will have to improvise. In this environment, lessons learned by one crew could be very important for subsequent crews, but just as easily could be lost — not just because details could be omitted in crew reports, but because new crew members cannot be briefed on every contingency that happened to every previous crew. However, crew members make detailed reports, which are compiled into narratives on the MDRS web site and distilled into future versions of the operations manuals. While there is no substitute for reading these materials prior to hands-on training by experienced crew members, there nevertheless exists an opportunity to develop information support systems that could deliver this knowledge when the crew needs it.

I propose to perform a human factors assay to determine the feasibility of developing an interactive crew advisory system suitable for deploying in hab, on an ATV or on a personal digital assistant. Mission control and previous crews have built up a large body of knowledge potentially useful to crews in the form of manuals and mission reports, and a combination of information retrieval and case based reasoning approaches could easily make this information available when the crew confronts problems in the field; however, the feasibility of this approach depends crucially on the patterns of work of the crew on site, the kinds of questions they naturally encounter in the field, the relevance of compiled data sources to these questions, their willingness to use computer assistance to help perform their tasks, and their buy-in on incorporating their own lessons learned into the system to support future crews. Crew buy in is critical in order to build a growing body of knowledge that will be of use to future crews, and therefore it is highly important to collect data about actual crew work patterns to inform the development of the tool.

Relevant Experience and Technologies

I have worked on two systems relevant to the proposed tool: the Captain's Advisory Tool and Nicole-IRIA.The Captain's Advisory Tool was a case-based reasoning system I developed at Georgia Tech with Kenneth Moorman that gave advice to a hypothetical spacecraft commander based on its case libraryof previous encounters. While this was a toy system, some of the ideas of CAT were later used in the Nicole-IRIA system, an information retrieval engine I prototyped at Georgia Tech and later participated in the commercialization of at Enkia Corporation. The core artificial intelligence idea of the Nicole series systems was to use the current reasoning context to guide memory retrieval. However, one key aspect that became clear during the commercialization phase is that it is critical to package these ideas in a way that ensures that:

• available content is useful to a user and the user's task
• system interface can integrate into the user's work patterns
• overall system functionality and workflow are packaged in a way that promotes user buy-in

If these conditions are achieved, there are a variety of artificial intelligence, information retrieval and groupware technologies that can be used to rapidly develop a prototype that will meet the end user's needs. The trick is collecting enough data to get a clear picture of how to meet the above demands. I propose to observe and document the actual crew work patterns and information resources available at the MDRS as a way to start to answer those questions, and to establish observation protocols we could use to extend the data collection to future crews.

On-Site Activities
On site, this project will:

• conduct a survey of the information resources available at the Hab
• collect observations of work patterns crewmembers performing common tasks:
• attempt to identify the information needs generated by crewmembers
  (to constrain what information will be useful in the proposed system)
• identify the physical contexts in which information needs arise
  (to identify the formfactors of potential assistance devices and interfaces)
• identify existing information resources that meet those information needs
• interview crewmembers about whether the resources meet those needs
• if possible, perform rapid prototyping of technology candidates, e.g., a search engine over a wiki

Questions of Interest
Some questions that might be addressed during this survey include:

• What on-hand information resources do crew members actually use during mission operations?
• What information needs do crew members generate?
• What are the typical needs satisfied by training or resources at hand?
• What needs require disrupt workflow with lookup in resources not at hand?
• What needs are not met, but crew members "work around"?
• What "lessons learned" are generated from typical crew activities?
• How are these lessons learned captured?
• Is there tacit knowledge which crews repeatedly have to re-learn?
• What unmet needs map to resources available or lessons learned from previous crews?

Post-Mission Deliverables
The products of this project will be:

• produce an inventory of information resources and their usefulness
• report initial findings of information needs, work patterns and potential form factors
• generate standardized observation recording forms for data collection on subsequent crews
• requirements / design assessment of proposed crew advisory tool

If the results of the initial findings are favorable, we will request that future crews conduct similar human factors observations and assessments using the observation recording forms generated during this mission. With data collected from several crews, we can start to address the questions of the value of past experience to future crews, and establish a profile for the advisory tool's information content, and outline the form factors and workflow requirements for crew members to buy in to use of the tool.

Selected References

Francis, A.G., Devaney, M., Ram, A. & Santamaria, J.C. (2001) Scaling Spreading Activation for Production Information Retrieval. In Arabina, H.R., (Ed.) Proceedings of the International Conference on Artificial Intelligence IC-AI’2000, Las Vegas, Nevada, June 26-29, 2000. CSREA Press.

Francis, A.G. (2000) Context-Sensitive Asynchronous Memory: A General Experience-Based Method for Managing Information Access in Cognitive Agents. Doctoral dissertation. College of Computing, Georgia Institute of Technology

Ram, A. & Francis, A.G. (1996) Multi-plan adaptation and retrieval in an experience-based agent. In David Leake, Ed., Case-Based Reasoning: Experiences, Lessons, Future Directions. MIT Press.