Task Progress:
|
NOVEMBER 2015 FINAL REPORT INFORMATION:
The overall objective of the AD ASTRA project (Automated Detection of Attitudes and States through Transaction Recordings Analysis, NRA grant #NNX12AB40G) was to identify suitable combinations of automated text processing techniques (which we call “Non-Intrusive Psycho-Social State Assessors” or NIPSSAs) and data streams for assessing psycho-social states of interest to NASA and to validate our overall hypothesis that NIPSSA measures could replace survey data for assessing selected psychosocial states. It had long been known that astronauts’ journals, diaries, and blogs were a rich source of material about their experiences and attitudes toward those experiences, and interpersonal communications are, after all, one of the best methods humans have of assessing changes in attitudes and states. We reasoned that if we could automate assessments of such data streams—many of which are already being captured and, to some extent, monitored during space missions—we would greatly expand the set of data which NASA team dynamics researchers have available which would help to close multiple Team and B-Med gaps and potentially improve the psychosocial health of future astronauts. Better, we could do so while concurrently improving astronauts’ daily lives and productivity by reducing the volume of surveys they are asked to complete.
The first year of our project surveyed the available techniques and data streams, arrived at the prioritization of textual analysis of journal/log entries and of interactive task discourse (in email, chat, and transcribed speech), and then went on to demonstrate promising sensitivity in the tools via analysis of pre-existing, historical data sources. While we showed individual and temporal variations with anecdotal correspondence to events, no independent data about the writers’ emotional states existed for validation of our text analysis techniques. The second year was targeted at transitioning the promising NIPSSA techniques we found with the first year’s work on historical data to validation studies with newly collected “live” data from ongoing experiments in analog environments (particularly the FARU Bed Rest Study Campaign 11 conducted at the Flight Analog Research Unit at University of Texas Medical Branch (UTMB)).
The third project year (which ended in August, 2014) was devoted to completing the bed rest analog journal study we began in Year 2, continuing to participate in a 4-month crew habitat study in the HI-SEAS (Hawaii Space Exploration Analog & Simulation) analog and beginning participation in an 8-month HI-SEAS study collecting both journal and crew-ground textual interaction messages (with corresponding survey data), and finally, developing techniques for audio data collection and then transcription in the HERA (Human Exploration Research Analog) analog, in which we obtained and analyzed journal, survey, text and speech interaction data from four separate week-long missions.
In our fourth and final, partial year (through February, 2015), we completed analyses of our HERA transcription data, and performed some cross-analog analyses illustrating the robustness of our techniques, and produced our Final Report. We were asked to take a 6 month no-cost extension by NASA personnel (through August, 2015) in which we provided a final outbriefing to NASA, prepared and submitted our data, and produced several articles. In all three analogs, we collected participant journals and daily surveys to cross validate our automated assessment techniques. In the HERA study, we also collected more than 160,000 minutes of captured audio recordings across four mission and transcribed large portions of this to enable examination of automated processing of interactive speech data.
The results have shown that automated textual analysis of subjects’ free form journals can reliably replace at least some forms of attitude and emotion surveys since our automated journal analyses regularly correlated significantly with subjects’ own survey ratings of their emotional positivity/negativity, and their focus on past/present/future across all three of our analog studies. Correlations between their attitudes about the study and about their physical well-being, as well as their focus on themselves vs. others, were also repeatedly significant. Furthermore, there is evidence that automated evaluation of journal entries may do a better job of assessing participant attitudes than do daily observers: a comparative analysis between our automated assessments and subjects’ own survey responses vs. the survey assessments of the subject by a nurse showed that automated analyses did a better job at predicting subjects’ survey responses.
We also demonstrated the speed and flexibility of our approaches by rapidly producing multiple after-the-fact analyses of elements of participants’ attitudes and emotions throughout the study including analyses of the effects of exercise and of testosterone treatments on subjects’ emotions and attitudes about sleep and food. Further, by examining the correlates of positive or negative mentions of a word group (e.g., food, eating, specific food items) we can provide suggestive evidence about how a specific individual or group thinks about that topic—e.g., that food is comforting and tied to nostalgia and social contacts or that it is a source of concern about body image and health. This flexibility to perform post hoc and deep, individualistic analyses (frequently in support of other researchers) proved to be a particularly useful capability and a strong distinguisher from survey questions.
In addition to journals, we developed techniques for examining interactive dialogue during tasks and daily activities. This form of communication is more immediate and less pre-meditated than journal entries and, therefore, is expected to be a less guarded but more transitory source of attitudes and emotions. We have detected significant differences in the overall emotional valence of the different crews, of valence for differing tasks among the different crews, and of varying (and differing) drivers for those emotional sentiments among the crew members.
During debriefs with insightful analog participants and operations personnel and throughout the 3.5 years of this program, we have identified a number of recommendations for the use of the developed techniques that are particularly relevant to long duration missions. Among these are an intervention or advisory tool for crewmembers themselves (reflecting exhibited changes in mood or attitude), a “power level” and “team comfort/routine” indicator for tracking intra-crew dynamics, and an aid for ground researchers to glean attitude and topic focus information from mission and training debriefing sessions. As recognized by the Behavioral Health & Performance (BHP) element, it will be crucial to provide accurate and low cost detection methods to detect measures relevant to BHP’s Team, BMed, and Sleep and Fatigue risks. We present data supporting the validity of ADASTRA techniques as one such approach.
SEPTEMBER 2014 REPORT: The overall objective for the project is to identify suitable combinations of processing techniques (which we call “Non-Intrusive Psycho-Social State Assessors” or NIPSSAs) and data streams for assessing psycho-social states of interest to NASA. The second year of our project was targeted at transitioning the promising NIPSSA techniques we found with the first year’s work on historical data to validation studies with newly collected “live” data from ongoing experiments in analog environments. The third project year has been devoted to completing the bed rest analog journal study we began in Year 2, continuing to participate in a 4 month long crew habitat study in the HI-SEAS analog collecting both journal and crew-ground textual interaction messages (with corresponding survey data), developing techniques for audio data collection and then transcription in the HERA analog, and then obtaining and analyzing journal, survey, text, and speech interaction data from HERA missions. In further testimony to the ease, speed, and flexibility with which data can be analyzed using our tools, we have been able to conduct novel analyses examining topics not planned for initial surveys using the bed rest journal data, and we have been among the first researchers to provide analyses of HERA crew data after each mission.
Significant accomplishments from the bed rest analog include:
1. Continued work in the bed rest analog allowed us validate several of our general techniques against survey data. These include: a. Significant correlations between our scoring of use of positive and negative emotional terms and the corresponding survey scores on the PANAS (Positive and Negative Affect Schedule) instrument. b. Significant correlations between the use of past, present, and future verb tenses in journal entries and a self-report of focus on past, present, and future. c. Significant positive correlation between mentions of a category of terms pertinent to the bed rest study in journal entries and ratings of attitude toward the survey. d. Significant negative correlation of use of terms in a category corresponding to physical state and the survey ratings of the participant’s own physical state.
2. We developed novel techniques and extensions for assessing the “drivers” of attitude on an individual by combining word category use with sentiment analysis. This let us draw conclusions, for example, that one individual’s use of food terms in journal entries with negative affect correlates with his use of body, physical, and health terms—which we interpret as likely indicating concerns about what the hospital diet is doing to his physique—while another uses food terms with positive affect in entries where he also uses past tense verbs and terms for social activities and family—which we interpret as likely indicating that he associates food with nostalgia for social gatherings and events. This type of insight could greatly help ground support in understanding how a psychological intervention or incentive might be interpreted by various crew members.
3. We identified an overall “positivity metric” with a potential “red line” threshold for adverse positivity expressions in journal entries, based on the work of Losada (Losada and Fredrickson, 2005), which suggests that a ratio of positive to negative utterances around 3:1 is indicative of a “thriving” individual or relationship. While this work is controversial, even discredited to some extent, and should be treated with caution, results did seem promising in our data, see below.
4. In addition to our planned analyses, we illustrated the speed and flexibility of applying our techniques (as well as the richness of journal entries as opposed to survey questions) by performing multiple auxiliary analyses which had not been anticipated in our initial experimental plan—such as on the psychological effects of exercise and testosterone treatments. These analyses were typically done in a 1-3 days of person time given our prior data set.
In addition to these validation analyses, we also were able to draw some substantive conclusions about the psychological states and the use of journals in long duration confinement conditions (for which the bed rest conditions serve as an analog). These include:
1. Although there are large individual differences, including some which countered this trend, word count of journal entries declines significantly over time.
2. Within that general trend, the proportional use of positive emotion terms, affect terms in general, perceptual terms (indicative of things being noticed or sensed in the environment), and cognitive mechanism terms (expressions of thought and reasoning) all decline significantly over time. Future tense verb use, though, increased significantly over time.
3. The analysis of positivity ratios described above showed that these ratios were correlated with trait anxiety surveys in our study and were also correlated positively with the use of terms in achievement and work categories in journal entries, and negatively correlated with use of terms indicative of health, physical state, anxiety, anger, frustration, and discouragement. Furthermore, the one subject whose ratio was below that 3:1 threshold in our sample was one who dropped out of the study.
4. Our analysis of the journals of exercisers vs. non-exercisers vs. exercisers also treated with testosterone showed:
a. The overall affect in journal entries (as measured by LSA valence) is higher for exercisers (in both groups) than non-exercisers. Exercisers rate (on surveys) their physical state as higher and they use physical and body terms more frequently, and with higher valence. Their valence for terms having to do with exercise is also higher.
b. When we discriminated between exercisers and exercisers with testosterone, however, we saw that exercise with or without testosterone clearly has substantial benefits on attitude. There was, however, no obvious added benefit from testosterone use on these fronts, over exercise alone. On the other hand, exercisers who did not have testosterone gave more positive survey responses (as opposed to our automated inferences from their journals) on explicit questions about several attitudes or behaviors than did exercisers with testosterone (who did not differ significantly from controls). Of somewhat more concern, testosterone subjects had twice the level of anxiety term usages in their journal entries as did either exercisers without testosterone or controls.
5. We did a comparative analysis between our automated assessment of subject journals and subjects’ own survey responses vs. the survey assessments of the subject by a nurse who interacted with him throughout the day. These results showed that our automated analyses did a much better job at predicting subjects’ survey responses than did the nurses’ survey answers.
While we have completed the bed rest study of subject journals during this program year, we have also been participating in two additional studies the HI-SEAS (Hawaii Space Exploration Analog & Simulation) and HERA (Human Exploration Research Analog) habitat analogs. In neither case are analyses from these studies complete, but we have been able to develop and extend our techniques in methodological ways to enable data capture. These include:
• In the HI-SEAS study, we have collected four months of both journal and interactive email-like exchanges between crew and ground. We developed extended survey techniques to obtain data to validate analyses of inter-crew and crew-ground “familiarity” or social distance, the degree of routine vs. unusualness of a day, etc. We have also now adapted our interpersonal communications analysis techniques for power and team comfort/routine to be able to handle the data from the BaseStation software used by the HI-SEAS team to collect and sequence crew-ground communications. We have now completed an initial run of our analytic software on both journal and crew-ground interactions, but have not yet completed analysis of the results.
• We extended those survey and data collection techniques for use in the data and communications environment of the week long studies in the HERA analog. More extensively, we researched, worked with HERA facility personnel, tested, developed, and eventually deployed an approach to capturing spoken interactions between crew and from crew to ground in the analog. This approach has now been used in the first four HERA missions.
• We also developed techniques for transcribing the captured audio from HERA missions in a cost-effective fashion. This represents a substantial effort in the past 6 months as we have explored automated, semi-automated, and human transcription setups, then consulted with and evaluated multiple transcription services, and ultimately identified a specific service provider and developed a process for selecting portions of the ~40,000 minutes of audio recording produced per HERA mission which were both needed by us and the other researchers in the HERA studies, and then transcribed, formatted, and shipped those portions.
The results achieved during this program year now demonstrate both the viability and the validity of one class of non-intrusive psycho-social state detectors to streamline, speed and in some cases, enable, the collection of data to effectively monitor and measure team and individual health and performance fluctuations during autonomous, long duration exploration missions. For journal data, we have been successful at applying NIPSSA techniques to a variety of “live” analog settings and have shown that they correlate with traditional survey measures. We have also shown that such data is far more flexible and rich than traditional survey measures—enabling rapid, later analysis for questions not articulated or of interest during initial study development. Finally, somewhat beyond initial expectations, we are supplying data relevant to retiring BHP gaps using these NIPSSA techniques. We have designed techniques and are collecting and analyzing data from additional analog studies where we are collecting interpersonal task communications data in text and speech to enable further detailed analyses of team dynamics using similar and alternate NIPSSA techniques.
Reference: Fredrickson BL, Losada MF (2005). "Positive affect and the complex dynamics of human flourishing." Am Psychol 60 (7): 678–86.
|
|
Abstracts for Journals and Proceedings
|
Miller C, Schmer-Galunder S, Wu P, Ott T, Rye J. "Non-Intrusive Psycho-Social State Detection for Attitudes with Exercise." 2014 NASA Human Research Program Investigators’ Workshop, Galveston, TX, February 12-13, 2014. 2014 NASA Human Research Program Investigators’ Workshop, Galveston, TX, February 12-13, 2014. http://www.hou.usra.edu/meetings/hrp2014/pdf/3315.pdf , Feb-2014
|
|
Abstracts for Journals and Proceedings
|
Miller C, Schmer-Galunder S, Ott T, Wu P, Rye J. "Non-Intrusive Psycho-Social State Detection: Results from Bed Rest Journals." 2014 NASA Human Research Program Investigators’ Workshop, Galveston, TX, February 12-13, 2014. 2014 NASA Human Research Program Investigators’ Workshop, Galveston, TX, February 12-13, 2014. http://www.hou.usra.edu/meetings/hrp2014/pdf/3316.pdf , Feb-2014
|
|
Abstracts for Journals and Proceedings
|
Miller C, Fischer U, Smith-Jentsch K, Kozlowski S, Mosier K, Wu P, Whitmore M. "Research in Long Term Human Performance in Space: Methods and Implications." Panel at 58th Annual Meeting of the Human Factors and Ergonomics Society, Chicago, IL, October 27-31, 2014. 58th Annual Meeting of the Human Factors and Ergonomics Society, Chicago, IL, October 27-31, 2014. , Sep-2014
|
|
Abstracts for Journals and Proceedings
|
Vessy B, Kozlowski S, Miller C, Roma P, Salas E, Tannenbaum S. "The Use of Analog Environments for Team Research." 30th Annual Conference of the Society for Industrial and Organizational Psychology, Philadelphia, Pennsylvania, April 23-25, 2015. 30th Annual Conference of the Society for Industrial and Organizational Psychology, Philadelphia, Pennsylvania, April 23-25, 2015. , Apr-2015
|
|
Abstracts for Journals and Proceedings
|
Wu P, Schmer-Galunder S, Ott T, Miller C, Rye J. "A Grounded Theory Approach to Individual and Team Performance Using Quantitative Analysis of Qualitative Metrics." Presented at the 31st Annual Meeting of the American Society for Gravitational and Space Research, Alexandria, VA, November 11-14, 2015. 31st Annual Meeting of the American Society for Gravitational and Space Research, Alexandria, VA, November 11-14, 2015. , Nov-2015
|
|
Abstracts for Journals and Proceedings
|
Miller C, Wu P, Ott T, Schmer-Galunder S, Rye J. "Non-Intrusive Psyco-Social State Detection: Technique Validation and Multiple Applications." 2015 NASA Human Research Program Investigators' Workshop, Galveston, TX, January 13-15, 2015. 2015 NASA Human Research Program Investigators' Workshop, Galveston, TX, January 13-15, 2015. , Jan-2015
|
|
Abstracts for Journals and Proceedings
|
Miller C, Wu P, Schmer-Galunder S, Ott T, Rye J. "AD ASTRA Automated Detection of Attitudes and States through Transaction Recordings Analysis-- Techniques, Results and Opportunities." To be presented at the 2016 NASA Human Research Program Investigators' Workshop, Galveston, TX, February 8-11, 2016. To be presented at the 2016 NASA Human Research Program Investigators' Workshop, Galveston, TX, February 8-11, 2016. , Feb-2016
|
|
Articles in Peer-reviewed Journals
|
Salas E, Tannenbaum SI, Kozlowski SWJ, Miller CA, Mathieu JE, Vessey WB. "Teams in space exploration: A new frontier for the science of team effectiveness." Curr Dir Psychol Sci. 2015 Jun;24(3):200-7. http://dx.doi.org/10.1177/0963721414566448
, Jun-2015
|
|
Articles in Peer-reviewed Journals
|
Fischer U, Mosier K, Orasanu J, Morrow D, Miller C, Veinott B. "Exploring communication in remote teams: Issues and methods." Proceedings of the Human Factors and Ergonomics Society Annual Meeting. 2013 Sep;57:309-13. 57th Annual Meeting of the Human Factors and Ergonomics Society, San Diego, CA, September 30-October 4, 2013. http://dx.doi.org/10.1177/1541931213571068 , Sep-2013
|
|
Articles in Peer-reviewed Journals
|
Miller CA. (Panel Participants: Fischer U, Smith-Jentsch K, Kozlowski S, Mosier K, Wu P, Whitmore M.) "Research in Long Term Human Performance in Space: Methods and Implications." Proceedings of the Human Factors and Ergonomics Society Annual Meeting. 2014 Sep;58(1):72-6. http://dx.doi.org/10.1177/1541931214581016 , Sep-2014
|
|
Books/Book Chapters
|
Miller C, Rye J, Wu P, Schmer-Galunder S, Ott T. "Team PsychoSocial Assessment via Discourse Analysis: Power and Comfort/Routine." in "Social Computing, Behavioral-Cultural Modeling and Prediction. 7th International Conference, SBP 2014, Washington, DC, USA, April 1-4, 2014. Proceedings." Ed. W.G. Kennedy, N. Agarwal, S.J. Yang. Springer International Publishing, 2014. (Series: Lecture Notes in Computer Science. Volume 8393, p. 309-316.) http://dx.doi.org/10.1007/978-3-319-05579-4_38 , Apr-2014
|
|
Papers from Meeting Proceedings
|
Schmer-Galunder S, Wu P, Rye J, Ott T, Miller CA. "Towards an Index of Mental Wellbeing in Language: The relationship between time orientation, self-focus and mood during prolonged bed-rest through automated analysis of daily journals." Presented at the Fifth International Conference on Social Media Technologies, Communication, and Informatics (SOTICS), Barcelona, Spain, Nov 15-20, 2015. Proceedings of SOTICS 2015 : The Fifth International Conference on Social Media Technologies, Communication, and Informatics, 2015. p. 103-108.
, Nov-2015
|
|
Papers from Meeting Proceedings
|
Wu P, Miller C, Ott T, Schmer-Galunder S, Rye J. "Mining For Psycho-Social Dimensions through Socio-Linguistics." 2015 AAAI Spring Symposium Series, Palo Alto, CA, March 23-25, 2015. Sociotechnical Behavior Mining: From Data to Decisions? Papers from the 2015 AAAI Spring Symposium. March, 2015. p. 33-38. http://www.aaai.org/ocs/index.php/SSS/SSS15/paper/view/10250/10097 ; accessed 12/8/15.
, Mar-2015
|
|
Papers from Meeting Proceedings
|
Wu P, Schmer-Galunder S, Ott T, Miller C, Rye J. "Automated Linguistic Approach to Derive Affect, Attitudes, and Other Psycho-Social Dimensions from Narratives." Presented at 2015 Human Factors and Ergonomics in Health Care, Baltimore, MD, April 26-29, 2015. Proceedings of Human Factors and Ergonomics in Health Care, Baltimore, MD, April 26-29, 2015.
, Apr-2015
|
|
Patents
|
Serial Number 14/733,692. Continuation from patent 8,825,584, filed June-2015. Jun-2015 Miller CA, Wu P, Rye J, Funk H, Ott T, Schmer-Galunder S. "Systems and Methods for Determining Social Perception Scores."
|
|
Patents
|
Serial Number 14/742,373. Continuation from patent 8,825,584, filed June-2015. Jun-2015 Miller CA, Wu P, Rye J, Funk H, Ott T, Schmer-Galunder S. "Systems and Methods for Determining Social Perception Scores."
|
|
Patents
|
Serial Number 14/742,378. Continuation from patent 8,825,584, filed June-2015. Jun-2015 Miller CA, Wu P, Rye J, Funk H, Ott T, Schmer-Galunder S. "Systems and Methods for Determining Social Perception Scores."
|
|
Patents
|
9,053,421. Continuation (child patent) to 8,825,584 awarded June-2015. Jun-2015 Miller CA, Wu P, Rye J, Funk H, Ott T, Schmer-Galunder S. "Systems and Methods for Determining Social Perception Scores."
|
|
Patents
|
8,825,584. Patent awarded, Sept 2014. Sep-2014 Miller CA, Wu P, Rye J, Funk H, Ott T, Schmer-Galunder S. "Systems and Methods for Determining Social Perception Scores."
|
|