=Paper=
{{Paper
|id=Vol-478/paper-5
|storemode=property
|title=Group Situational Awareness: Being Together
|pdfUrl=https://ceur-ws.org/Vol-478/paper5.pdf
|volume=Vol-478
|dblpUrl=https://dblp.org/rec/conf/um/DimK09
}}
==Group Situational Awareness: Being Together==
https://ceur-ws.org/Vol-478/paper5.pdf
Group Situational Awareness: Being Together
Eyal Dim and Tsvi Kuflik
The University of Haifa, Mount Carmel, Haifa, 31905, Israel
dimeyal@cri.haifa.ac.il . tsvikak@mis.haifa.ac.il
Abstract. In many cases, museum visitors come to the museum in small groups of friends or families. Their
level of ‘togetherness’ may be implied by their proximity and interaction. Position proximity is a basic
requirement to enable quiet face to face conversation in a museum, while voice communication is an example
of interaction. Group ‘togetherness’ may be measured to serve two purposes: (1) on the micro level,
identifying if the group members are currently together or apart, and (2) on the macro level, identifying group
characteristics (such as cohesion). This study focuses on the micro level in a museum environment,
presenting observations and analysis that intend to set the foundations for automatically measuring and
analyzing ‘togetherness’ among museum visitors.
Keywords: Group Interaction, Group Model, Cultural Heritage, Users Study.
1. Introduction
The museum world is looking for innovative technologies which may enhance their visitors' experience.
Bitgood [2] posits that overwhelming percentage of museum visitors come in groups. He sees the social contact
as a very important aspect of informal learning settings, and sometimes as the most important part of the
museum visit experience. Interaction between visitors is known to enhance the museum experience, deepen the
visitors’ involvement and increase the intimacy among group members [14]. When people visit the museum in
small groups of family or friends, the social context is different from the case of an individual visitor [4]. In the
group visit case people share their attention between the exhibits or the guidance (such as labels, audio guides,
handheld computers, etc.) and the other group members.
There are several questions that need to be answered: (1) what measurements can a system use to be aware of
the situation of the group and its members? (2) can a technology be aware of higher level group characteristics
(such as ‘group cohesion’)? (3) interrupt management question: when should an intelligent mobile device
intervene during the museum group visit? and (4) what kind of intervention may such a technology apply? This
work focuses on the first question in the light of the others, by trying to measure the group ‘togetherness’(the
term ‘togetherness’ refers to a social activity which enables mutual sharing of thoughts, feelings, knowledge,
wants and needs, among group members), as a required pre-condition for interaction.
There may be several levels of ‘togetherness’ that a group has for its members and that the members have for
one another [5]. Therefore, understanding group ‘togetherness’ may allow using technology to better support
both the group needs and its members’ interests. This study identifies social-interaction as a measure that may be
used to identify group ‘togetherness’. It measures physical proximity as a pre-condition for face-to-face social
interaction and conversation. A group may be together or apart, and its group members may join or leave the
group. Assuming that the group members use an intelligent mobile device, the application may, for example,
provide recommendations to a group member when he/she is not involved in deep conversation with others (free
attention); or it may adapt to the group and its members by making recommendations for those group members
who are close to each other, neglecting the separated group members interests. If the application has, for
example, information about the level of group cohesion, it also may choose to treat the group differently.
Previous studies focused mainly on exploring the possibility to use novel technologies to support individuals
visiting the museum, mainly by improving the ways of information delivery [8]. This included adaptation [13],
personalization [11] and various additional aspects such as context awareness [3], support of positioning and
navigation [9], and visitors’ circulation [7]. Several studies dealt with recommendations that may suit most small
group members, based on a variety of strategies [6]. Some applications such as Sotto Voce [1], ARCHIE [12],
PEACH [16], PIL [10] and AgentSalon [17] were aimed at using collaborative tools like messaging, voice
communication and eavesdropping to enable intra group interaction.
This work focuses on basic measurements that may enable group situational awareness in a museum visits
setting. It intends to evaluate the possibility of automatically measuring and analyzing the group ‘togetherness’
based on proximity and interaction of museum visitors. These measurements, in turn, may be used to predict the
group behavior and trigger the adaptation of a system to meet the needs of the small group and its members.
Prediction of group behavior is important for better group monitoring and situational awareness. It may lead to
actions within the current applications or may be shared with other applications and contribute to enhance the
1
museum visitor's experience. Moreover, these initial results may be applied in different scenarios with similar
characteristics, such as cultural heritage sites and tourism in general, large exhibitions, and shopping malls.
2. Measuring Visitors Proximity as an indicator for Group ‘togetherness’
2.1. Setting and Data collection
Proximity and voice interaction data was collected at the "Yitzhak Livneh-Astonishment" exhibition
presented in the Tel-Aviv Museum of Arts, where 142 visitors were observed in 58 groups (Table 1), by their
random entry order. There was no human guidance at the exhibition hall, and all visitors of this exhibition did
not use any other guidance (written, audio, or other intelligent mobile device). This specific data was collected
since it represents obvious group ‘togetherness’ behaviors. Proximity of group members had one of three states
(which was the dominant state during the sampling interval of 1 minute): (i) Separated – all group members are
separated (at least two exhibits apart or two meters apart). (ii) Joined – Some group members are together. (iii)
Left – All group members left the exhibition. On the average, groups were “Separated” 30.2 percent of the
observed visit time. The duration of Voice interaction was recorded within each sampling interval of the
observation (1 minute). Voice conversation provides better evidence of interaction. Data was collected by
observations, but in the future such data may be collected by technologies such as the wearable sociometric
badge, which collects location, proximity, orientation, human activities and speech features data [15].
Table 1. Summary of group's characteristics
# of Groups Observed Group Gender
Group
Size Family Family - Males Females
Couples Friends Mix
(Children under 18) Other Only Only
2 23 2 5 10 1 15 24
3 N/A 6 1 4 0 2 9
4 N/A 3 0 3 0 1 5
5 N/A 1 0 0 0 0 1
2.2. Proximity Measurements Analysis
The sampled proximity data created proximity patterns, represented as the vectors of “Joined” and “Separated”
states over time. Each vector element has a state for every relevant time step (1 minute), and the state vector has
10 elements (10 observation minutes). The position-proximity patterns can be used to describe the level of
‘togetherness’ of group members, based on a criterion to decide what it means to be together and what it means
to be apart during the group visit. The criterion suggested here is a separation ratio (or its complementary
criterion - a join ratio), which operates on the group. Let J be the number of “Joined” periods along the group
visit (within the proximity patterns), S be the number of “Separated” periods along the visit, and let JR be the
"Join Ratio" and SR be the "Separation Ratio" then: JR is defined by equation 1 and SR by equation 2. Of course
this leads to equation 3:
(1) JR=J/(J+S) .
(2) SR=S/(J+S) .
(3) SR=1-JR .
This definition enables reorganization of the “Joined” and “Separated” position proximity patterns as shown in
Table 21. Columns 2 through 11 are the minutes of measurement – the cells contain the value of "1" for “Joined”
states and the value of "2" for "Separated” states while "0" is used for minutes when the group already "Left" the
exhibition. Column 12 presents the SR value and column 13 presents the JR value. Columns 14 through 16
respectively show the “Joined” state-count, the “Separated” state-count and their totals. The proximity patterns
generated by the “Joined” and “Separated” vectors have been sorted first by the SR value and then by the time
the group was present at the exhibition (equals the total of “Joined” and “Separated” minutes as presented in
column 16).
1
Due to space limitations, only sample of the patterns are presented
2
Table 2. Adaptation decision based on position-proximity patterns
Joined Separated Total
Group ID Minute 1 Minute 2 Minute 3 Minute 4 Minute 5 Minute 6 Minute 7 Minute 8 Minute 9 Minute 10 SR JR Count Count Count
29 1 1 1 1 1 0 0 0 0 0 0.00 1.00 5 0 5
46 1 1 1 1 1 1 0 0 0 0 0.00 1.00 6 0 6
49 1 1 1 1 1 1 1 1 1 0 0.00 1.00 9 0 9
48 2 1 Joined zone
1 1 1 1 1 1 1 1 0.10 0.90 9 1 10
53 1 1 1 1 1 1 1 1 2 1 0.10 0.90 9 1 10
35 1 1 1 2 1 1 0 0 0 0 0.17 0.83 5 1 6
57 1 1 1 2 1 1 0 0 0 0 0.17 0.83 5 1 6
43 1 2 1 1 1 0 0 0 0 0 0.20 0.80 4 1 5
4 1 1 1 1 1 1 1 1 2 2 0.20 0.80 8 2 10
21 1 1 1 2 2 1 1 1 1 1 0.20 0.80 8 2 10
44 1 2 2 1 1 1 1 1 0 0 0.25 0.75 6 2 8
40 1 1 1 1 1 2 2 0 0 0 0.29 0.71 5 2 7
30 1 1 1 1 2 2 1 1 1 2 0.30 0.70 7 3 10
54
11
1
2
1
1
In between zone
2
1
1
2
1
2
2
1
2
1
1
0
0
0
0
0
0.38
0.43
0.63
0.57
5
4
3
3
8
7
15 1 2 1 2 0 0 0 0 0 0 0.50 0.50 2 2 4
18 2 2 2 1 1 2 1 1 2 2 0.60 0.40 4 6 10
55 1 1 2 2 2 2 0 0 0 0 0.67 0.33 2 4 6
34 1 1 2 2 2 2 2 2 1 2 0.70 0.30 3 7 10
16 2 1 2 2 2 2 1 0 0 0 0.71 0.29 2 5 7
3 2 1 2 2 1 2 2 2 2 2 0.80 0.20 2 8 10
23 2 2 Separated zone
1 2 2 2 1 2 2 2 0.80 0.20 2 8 10
45 1 2 2 2 2 2 2 0 0 0 0.86 0.14 1 6 7
42 2 2 2 2 2 2 2 2 2 0 1.00 0.00 0 9 9
47 2 2 2 2 2 2 2 2 2 2 1.00 0.00 0 10 10
Table 2 is divided into three sections, based on SR (or JR) values: thresholds of SR=0.7 (JR=0.3) for being
apart and SR=0.2 (JR=0.8) for being together. These SR/JR thresholds have been selected only for
demonstration purposes and need further study to be properly adjusted. However, this example shows how
position proximity may be measured and then analyzed for gaining some insight about group ‘togetherness’.
Even though it seems that position proximity may help in understanding groups, what if group members in
close proximity do not interact at all? This is where voice interaction comes into play. The voice proximity is
based on a threshold for cumulative duration of conversation within each predefined time measurement period (a
minute in this case). For example, if the threshold is set to 10 seconds, a group having 15 seconds of
conversation within a minute, is considered “Joined” (15 > 10 seconds), and a group having 5 seconds of
conversation within a minute is considered “Separated” (5 ≤10 seconds). By setting such a Voice Duration
Threshold (VDT) we can select our definition for quantified interaction. In addition we exchange the position
proximity patterns above with voice proximity patterns, based on “Joined” / “Separated” states that were defined
by the VDT.
By measuring proximity we may infer interaction while by measuring voice data we can prove interaction. As
the VDT grows, groups are more “Separated”. Detailed investigation of the groups’ voice interaction reveals that
even a requirement for a VDT of 10 seconds is enough to significantly change the SR of a specific group and
transform it from position proximity “Joined” to voice proximity “Separated”. Table 3 exemplifies this change in
behavior. It presents information about 5 groups. Each cell presents the “Joined” or “separated” state based on
proximity to the left of the arrow symbol (“→”) and “Joined” or “Separated” state based on VDT of 10 seconds
to the right of the arrow symbol. “J” represents a “Joined” state and “S” represents a “Separated” state. Darker
cells represent minutes where the state changed. If the group left the exhibition the cell is blank. The collected
data shows that if the VDT is high enough (>5 seconds), any voice-based separation would also mean position-
based separation (i.e. the change when the VDT increases is always towards more separation). For example the
VDT changed the state of group 25 from “Joined” to “Separated” for the total time of presence at the exhibition.
Such a change in the determination of “Joined” and “Separated” states could affect the decisions about
‘togetherness’ or cohesion, if they were based on voice proximity “Separated” criterion rather than on position
proximity “Separated” criterion.
Table 3. The Change from “Joined” to “Separated” States for a VDT of 10 Seconds
Group Minute Minute Minute Minute Minute Minute Minute Minute Minute Minute
ID 1 2 3 4 5 6 7 8 9 10
21 J→J J→J J→J S→S S→S J→J J→J J→J J→J J→J
22 J→J J→J J→J J→J J→J
23 S→S S→S J→J S→S S→S S→S J→S S→S S→S S→S
24 J→J J→S J→J J→S J→J J→J J→J
25 J→S J→S J→S J→S J→S
This study tested several values for the VDT. Higher values of VDT (i.e. longer conversations) mean that the
group members are more occupied with sharing the visit experience with each other (only 3% of the
conversations didn’t relate to the museum). The selection of VDT=10 above considered the following: (1) it was
the first value to have only ‘J → S’ changes (no ‘S → J’ changes), higher VDT values kept the same transitions’
direction; and (2) it was close to the position-proximity measurements, which means that conversation-proximity
may replace the position proximity as a measure.
3
2.3. Prediction of Behavioral Patterns
Being able to predict visitors’ behavior may allow selection of a course of action, hence the question presented
here is: how can the position-proximity pattern from the last several minutes be used to predict the position
proximity in the next minute? The data gathered during the observations is used to assess the feasibility of
prediction. Table 4 presents an example which summarizes all the sequences of 4 consecutive minutes. The
observed state in the 4th minute is compared with the pattern of the preceding 3 minutes. For 3 minutes we have
8 options of patterns (comprised of 3 components selected from a {Joined, Separated} set) starting from "Joined-
Joined-Joined", and ending with "Separated-Separated-Separated". These patterns appear in columns 1 through 3
of the table. Each row presents a different pattern. In column 4 we have the number of cases, where the pattern
on the left was followed by a “Joined” minute. In column 7 the results are presented in percentage with the same
three preceding minutes. In column 5 we have the number of cases where the pattern on the left was followed by
a “Separated” minute. In column 8 the results are presented in percentage with the same three preceding minutes.
Column 6 presents the total number of cases with the same three preceding minutes. The bottom row presents the
total number of cases analyzed, showing that even for the conservative measurement that we used (groups were
considered “Joined” even if only a sub-group was together), still 33% of the time groups were “Separated”.
We can cluster the results in Table 4 into four major categories: (i) All three “Joined” preceding minutes in the
pattern are the same (i.e. "Joined-Joined-Joined") – in this case the probability is high (≥90%) that the next
minute would be the same. (ii) All three “Separated” preceding minutes in the pattern are the same (i.e.
"Separated-Separated-Separated") – in this case the probability is quite high (≥79%) that the next minute would
be the same. (iii) The three minutes in the patterns are alternating between “Joined” and “Separated” (i.e.
"Joined-Separated-Joined" or "Separated-Joined-Separated") – in this case again, the probability is quite high
that the states in the first and third minutes repeat in the next minute (≥83%). (iv) In the four additional cases the
probability is not conclusive. It should be noted that the majority decisions for cases (iii) and (iv) are the same
while the prediction probability is totally different. The interpretation is that a consistent group in cases (i) and
(ii) would probably continue its behavior for the next minute. A group which deviated for a minute and returned
to its previous behavior would probably continue with that behavior for the next minute, as in case (iii), and a
group that changed its position proximity and kept it for the next minute would be unpredictable. Please note,
that in all cases, if the first and the last minute of the three minute sequence are the same the probability is high
(≥87%) that the 4th minute would be the same.
Table 4. Actual results of the next minute position proximity compared to the preceding three minutes position peoximity
Next Minute Actual Next Minute Actual Results
Previous 3 Minutes Pattern Results Total Percentage
Joined Separated % Joined % Separated
Joined Joined Joined 121 14 135 90% 10%
Joined Joined Separated 9 13 22 41% 59%
Joined Separated Joined 10 2 12 83% 17%
Joined Separated Separated 8 11 19 42% 58%
Separated Joined Joined 11 6 17 65% 35%
Separated Joined Separated 1 8 9 11% 89%
Separated Separated Joined 9 6 15 60% 40%
Separated Separated Separated 7 26 33 21% 79%
Total 176 86 262 67% 33%
This is an example of a possible analysis. Future analyses (using tools such as the Hidden Markov Model [18])
may assess the contribution of various lengths of proximity patterns history and the impact on the prediction of
the next minute. Other variables, such as the location of the group in relation to the exhibits, may also have an
impact on the prediction.
3. Conclusions
This work focused on the possibility to use technology for tracking small groups ‘togetherness’ in a museum
environment. Proximity hierarchy has been shown: position proximity is a precondition for voice proximity.
Position proximity and voice proximity patterns can serve as criteria for group 'togetherness' or even group
cohesion. The proximity patterns are group related aspects that may be measured and monitored automatically
by available technology. These measurements, in turn, may be used to predict the group behavior and trigger the
adaptation of a technology to meet the needs of the small group and its members. Prediction is important for
better group monitoring and improved situational awareness.
Acknowledgements. The work was supported by the collaboration project between the Caesarea-Rothschild
Institute at the University of Haifa and FBK/irst in Trento and by FIRB project RBIN045PXH.
4
References
1. Aoki P.M., Grinter R.E., Hurst A., Szymanski M.H., Thornton J.D. and Woodruff A., Sotto Voce: Exploring the
Interplay of Conversation and Mobile Audio Spaces. CHI 02, Proceedings of SIGCHI Conference on Human Factors.
In: Computing Systems, New York, NY, USA, ACM Press, 431-438 (2002)
2. Bitgood Stephen: Environmental Psychology in Museums, Zoos, and Other Exhibition Centers. In: R. Bechtel & A.
Churchman (eds.), Handbook of Environmental Psychology, John Wiley & Sons., 461-480 (2002)
3. Davies N. Cheverst K. Mitchell K. Efrat A.: Using and Determining Location in a Context-sensitive Tour Guide.
Computer 34, 8, 35-41 (2001)
4. Falk J.H., Dierking L.D.: The museum experience. Whalesback Books, Howells House, Washington (1992)
5. Johnson, W. David, Johnson, F.P.: Joining Together: Group Theory and Group Skills. Pearson Education, Inc. 9th edn.
(2006).
6. Kay J., Niu W.: Adapting Information Delivery to Groups of People. Proceedings of the First International Workshop on
New Technologies for Personalized Information Access, at the Tenth International Conference on User Modeling,
Edinburgh (2005)
7. Klein Hans-Joachim: Tracking Visitor Circulations in Museum Settings. Environment and Behavior, Sage Publications
(1993)
8. Kray Christian and Baus Yörg: A Survey of Mobile Guides. In conjunction with Fifth International Symposium on
Human Computer Interaction with Mobile Devices and Services (2001)
9. Krüger Antonio, Baus, Jorg, Heckmann Dominikus, Kruppa Michael and Wasinger Rainer: Adaptive Mobile Guides.
The Adaptive Web, LNCS 4321, Springer, Berlin Heidelberg New York, 521-549 (2007)
10. Kuflik T., Stock O., Zancanaro M., Sheidin J., Jbara S., Goren-Bar D., Soffer P.: Supporting Small Groups in the
Museum by Context-Aware Communication Services. Proceedings of the 12th International Conference on Intelligent
User Interfaces, Honolulu, Hawaii, USA, 305-308 (2007)
11. Kuflik Tsvi, Busetta Paolo, Penserini Loris, Bresciani Paolo and Zancanaro Massimo: Personalized Information
Delivery in Dynamic Museum Environment by Implicit Organizations of Agents. In Workshop on Environments for
Personalized Information Access, 22-33 (2004)
12. Luyten K., van Loon H., Teunkëns D., Gabriëls K., Coninx K., and Manshoven E.: ARCHIE: Disclosing a Museum by a
Socially-aware Mobile Guide. The 7th International Symposium on Virtual Reality, Archaeology and Cultural Heritage,
VAST 2006, M. Ioannides D. Arnold, F. Niccolucci, K. Mania (Edts.) (2006)
13. Masthoff Judith: Selecting a Sequence of Television Items to Suit a Group of Viewers. User Modeling and User-
Adaptive Interaction vol. 14(1), 37-85 (2004)
14. McManus P.: Towards Understanding the Needs of Museum Visitors. In B. and G.D. Lord (eds), Manual of Museum
Planning (London: HMSO), 35-51 (1991)
15. Olgu´ın Olgu´ın Daniel, Waber Benjamin N., Kim Taemie, Mohan Akshay, Ara Koji, and Pentland Alex
(Sandy): Sensible Organizations: Technology and Methodology for Automatically Measuring
Organizational Behavior; IEEE Transactions on Systems, Man, and Cybernetics—Part B: Cybernetics, vol.
39. no.1, (2009)
16. Stock Oliviero, Zancanaro Massimo, Busetta Paolo, Callaway Charles, Kr¨uger Antonio, Kruppa Michael, Kuflik Tsvi:
Not Elena, and Rocchi Cesare: Adaptive, Intelligent Presentation of Information for the Museum Visitor in PEACH.
User Modeling and User-Adapted Interaction, Vol. 18(3), 257–304, (2007)
17. Sumi Y. and Mase K.: Interface Agents That Facilitate Knowledge Interaction between Community Members. Springer
Verlag (Cognitive Technologies Series) , 405-427 (2004)
18. Russell S. and Norvig P.: Artificial Intelligence a Modern Approach; 2nd edition, Pearson Education, Inc, New Jersey,
USA, (2003)
5