Conference PaperPDF Available

Predicting creativity in the wild: Experience sample and sociometric modeling of teams

Authors:

Abstract

Relationships between creativity in teamwork, and team members' movement and face-to-face interaction strength were investigated "in the wild" using sociometric badges (wearable sensors), electronic Experience Sampling Methods (ESM), the KEYS team creativity assessment instrument, and qualitative methods, in academic and industry settings. Activities (movement and face-to-face interaction) and creativity of one five-member and two seven-member teams were tracked for twenty-five days, eleven days, and fifteen days respectively. Paired-sample t-test confirmed average daily movement energy during creative days was significantly greater than on non-creative days and that face-to-face interaction tie strength of team members during creative days was significantly greater than for non-creative days. The combined approach of principal component analysis (PCA) and linear discriminant analysis (LDA) conducted on movement and face-to-face interaction data yielded a model that predicted creativity with 87.5% and 91% accuracy, respectively. Computational models that predict team creativity hold particular promise to enhance Creativity Support Tools.
Predicting Creativity in the Wild:
Experience Sample and Sociometric Modeling of Teams
Priyamvada Tripathi Winslow Burleson
School of Computing, Informatics, and Decision Systems Engineering
Arizona State University
{Pia, Winslow.Burleson} @asu.edu
ABSTRACT
Relationships between creativity in teamwork, and team
members’ movement and face-to-face interaction strength
were investigated “in the wild” using sociometric badges
(wearable sensors), electronic Experience Sampling
Methods (ESM), the KEYS team creativity assessment
instrument, and qualitative methods, in academic and
industry settings. Activities (movement and face-to-face
interaction) and creativity of one five-member and two
seven-member teams were tracked for twenty-five days,
eleven days, and fifteen days respectively. Paired-sample t-
test confirmed average daily movement energy during
creative days was significantly greater than on non-creative
days and that face-to-face interaction tie strength of team
members during creative days was significantly greater
than for non-creative days. The combined approach of
principal component analysis (PCA) and linear
discriminant analysis (LDA) conducted on movement and
face-to-face interaction data yielded a model that predicted
creativity with 87.5% and 91% accuracy, respectively.
Computational models that predict team creativity hold
particular promise to enhance Creativity Support Tools.
Author Keywords
Creativity Support Tools (CST), Sociometric Modeling,
Experience Sample Method, Wearable Computing.
ACM Classification Keywords
H.5.3 Group and Organization Interfaces
General Terms
Human Factors; Measurement.
INTRODUCTION
Understanding how to foster creative capacity is among the
most important goals of our society in preparation for the
future. While there are many definitions of creativity, there
is broad consensus that creativity is the creation of anything
that is useful and original [32]. Creativity takes place
through the unfolding of moment-to-moment activities in
natural environments. This investigation studies the
relationship between group activity characterized through
team members’ movement and face-to-face interactions
within teams, and creativity in research and development
teams in industry and academia. Group activity was tracked
through sensor data from sociometric badges [25]. A social
science survey instrument KEYS [2][5] was implemented
through electronic ESM and Day Reconstruction Method
[21] to capture self-reported creativity and was
supplemented with expert-coded creativity measures.
Statistical methods, machine learning, and qualitative
approaches were used to analyze and validate the
relationship between group activity (movement and face-to-
face interaction) and everyday creative and non-creative
events. This investigation contributes to the basic science
of creativity and to the empirical methodologies that assess
creativity. Providing movement and face-to-face
interactions as a means of continuously sensing creativity
accurately, with minimal human input, this research
contribute to the design of CST.
Three studies ranging in duration from two to four weeks
were conducted over a two-year period in leading research
and development laboratories. These laboratories had
similar spatial environments, consisting of a room with
cubicles facing walls and a central collaboration space. All
participants had regular working hours and the environment
encouraged hands-on, real time interaction towards
development of artifacts. To explore initial variables of
interest, a pilot study was conducted with five participants
over 25 days. Encouraged by the trends discovered in the
pilot study, two further experiments were conducted with
two, seven member teams, for 11 and 15 days, respectively.
The framework based on these studies paves the way for
automated CST based on team activity [47] [48]. This
framework combines existing qualitative methods of
studying group creativity in context [4] with sensor-based
data. The research was conducted in natural work place
settings and advances the use of sensor data to support
human creativity and teamwork.
BACKGROUND
Assessment of Creativity
Aimed at measuring creativity, several divergent thinking
tests have been designed that test for fluency or rate of
ideation in individuals. Some early tests proposed to test
creativity were Guilfords Structure of Intellect (SOI)
divergent production tests [19], and Torrances Tests of
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies are
not made or distributed for profit or commercial advantage and that copies
bear this notice and the full citation on the first page. To copy otherwise,
or republish, to post on servers or to redistribute to lists, requires prior
specific permission and/or a fee.
CSCW’12, February 1115, 2012, Seattle, Washington, USA.
Copyright 2012 ACM 978-1-4503-1086-4/12/02...$10.00.
Creative Thinking (TTCT) [51]. A limitation of these
parametric approaches is that they may measure only one
aspect of creativity and may only work in controlled
laboratory conditions. Alternate approaches have focused
on the creative qualities of the process and outcome or final
products, e.g., the creative product semantic scale [8] and
Amabile’s consensual assessment technique (CAT) [2].
CAT can be employed to address domain specific creativity
and also be extended for analysis of team creativity. CAT
relies on the inter-rater reliability between these expert
judges to evaluate levels of creativity in processes and
products. Amabile has also developed the KEYS scale
based on her componential model of creativity [3] [5] that
measures a variety of factors such as affect, rewards, and
motivation, and asks the participants to self report on
creativity (on a Likert scale) and to explain what they did
throughout the day in an open ended question format. This
scale addresses individual, team, and context variables
concurrently, with checks and cross checks to ensure
consistency and accuracy of responses.
Group Creativity Research
Work on creativity has largely focused on individual
creativity. However, in recent years, there has been an
increased interest in studying team creativity within
organizations. An important observation is that it is not
very clear how individual creativity is linked to team
creativity [43]. Researchers have often dealt with this
conundrum by focusing on individual creativity alone, or
on team creativity alone, or on processes of interactions
between team members [6][43][48][50].
Group creativity can be conceived as an additive or
disjunctive property of individual creativity [43]. If it is
additive, then each individual member’s creativity adds up
to the final creative output of the team. If it is disjunctive,
the most creative ideas which may come from one or more
individuals are adopted by the team. Team creativity may
also manifest itself as a weighted combination of individual
contributions. Amabile, in her componential model of
creativity, has pointed out that intra-individual factors such
as organization incentive for innovation, resources made
available, and external pressure can impact individual
creativity [2]. Taggar [50] found that in addition to
creativity relevant skills, domain relevant knowledge, and
intrinsic motivation at the individual level, there might be
group related processes that are relevant to creativity.
Amabile’s KEYS scale [5] quantifies such variables in a
reliable manner suitable for measuring team creativity.
Sensors for feedback and the display of group behavior
[13][25] have provided automated support for group
performance. These tools employ sensors in laboratory
environments and focus on single modality, such as speech
[28]. Several CSTs have also been developed that aim to
improve group creativity. Most of the studies in CSTs
center on the comparisons of an experimental group that
uses the CST system versus a nominal group (set of
individuals not acting as a group) or a control group that
does not use the system [31]. Studies tend to focus on
quantifying interactions and using them as a measure of
creativity and productivity.
Interactions and Movement in Group Creativity
Network communication strength and the types of
communication modalities within a team are additional
behavioral factors that may impact group creativity [26].
While there is a belief that face-to-face interaction strength
is central to the understanding of social networks in relation
to creativity, there is insufficient empirical evidence to
indicate strong relationship between face-to-face
interaction and creativity. Tie strength (weak, strong) is a
function of the amount and quality of interactions,
emotional intensity, and reciprocity that takes place
between two individuals [17]. Zhou et al. [55] found that
employees exhibited greater creativity when their number
of weak ties was neither too low, nor too high (an
intermediate level exhibited greatest level of creativity).
Perry-Smith and Shalley [42] showed that weak ties rather
than strong ties are beneficial for creativity among research
scientists. In contrast, Obstfeld [34] showed that engineers
with strong ties are more creative. These studies show a
complex, inconclusive, and possibly domain specific
relationship between tie strength and creativity.
Research has also investigated the nature of the structure of
social networks, within teams, that is most supportive for
creativity [35]. Burt [10] claims that a network with several
structural holes (many disconnected individuals) may be
more creative. Burt hypothesizes that members who are
closer to these structural holes are exposed to a greater
diversity of perspectives, which has a positive impact on
creativity. On the other hand, Perry-Smith and Shalley [42]
claim that a dense network with all members strongly
connected to each other provides an opportunity for free
interchange of information and hence greater creativity.
Most research conducted on the relationship between
movement and creativity is in the exercise sciences where
brief period of activity such as walking on the treadmill is
followed by creativity assessment questionnaires [30][49].
These studies have established that physical activity in
humans is linked to their creativity. However, research is
needed to understand how individual movement in the
work environment is related to creative production.
Sensor Based Approach to Group Creativity Research
Recent advancements in sensor based pattern recognition
with applications such as face recognition [53] and gait
recognition [27] demonstrate the power of detecting low
level signal streams and recognizing relevant patterns from
them. Sociometric badges [39], used to capture movement,
speech, and location, are an interesting recent example.
They are a form of environmentally aware computing
allowing capture of persons location, presence, and
elements of the environment [38]. HCI research has
leveraged such sensors using speech, and artifacts to study
interactions within members of groups [13][16][37].
Burleson et al.; Choudhary; Kapoor et al.; Kim et al.;
Olguin-Olguin et al.; and Pentland [9][11][22][25][37] [41]
have conducted empirical studies that use physiological
sensing and wearable computing to understand and predict
high level behavioral constructs such as affect, activity and
creative output. Methods from affective computing have
been able to distinguish affective state with 81% accuracy
throughout everyday activities [22] while machine learning
tools that incorporate Human Eigen behaviors and Coupled
Hidden Markov Models (CHMMs) have been shown to
account for 96% of the variance of behavior of typical
individuals [15].
Assumptions and Challenges in Analyzing Group Data
Group research is a challenging task [45]. The most
difficult challenge from a statistical perspective lies in
analyzing data from groups [20][24]. Data gathered from
individual participants in group studies is often
interdependent, which limits the use of several statistical
tools for analysis. Yet, employing many of these
techniques can be valid in group analysis as long as
observations are shown to be independent. A foremost
technique to show independence is correlation analysis [20]
[12][19][23][33]. In the investigations reported here,
calculation of intraclass coefficients confirmed the
assumption of independence of data to be valid.
INSTRUMENTS AND METHODS
The following instruments and methods were approved by
Institutional Review [52].
Sociometric badges: wearable sensors, worn as a pendant
around the neck, recorded network data (Infra Red pings) at
17 Hz, body movements (2D accelerometer) at 50 Hz and
ambient audio, not used in the present studies, using
embedded speaker at 8 kHz. Badges track location and
analyze elements of participant’s social interaction through
bi-directional infrared transceiver, accelerometer, and low-
resolution microphone analysis. No personally identifiable
data is recorded which ensures privacy of subjects. The raw
data from the sensors are extracted into meaningful features
that may correlate with team members’ characteristics.
Badges are equipped with triaxial accelerometers that give
the value of movement in X, Y, and Z directions. The mean
and standard deviation of movement energy for each
participant for each day was calculated. Movement energy
gives a measure of the intensity of individual movement
that includes the effect of variation in signal around the
three axes in the accelerometer [37]. It may also be
classified into various types of physical activity such as
walking, running, and sleeping [44].
Calculating Face-To-Face Tie Strength: Infrared signals in
the sociometric badges provide a measure of face-to-face
interaction. Badges record presence and duration of other
badges when they are in direct line of each other (IR signal
cone of height 1 meter and radius r h tan Ө where Ө =
±15º) [37]. We counted the number of pings for each badge
and constructed adjacency matrix for the data. Cells in the
adjacency matrix represent the number of pings recorded
for each badge with all other badges. This matrix was first
made symmetric with respect to the minimum number of
pings recorded for each pair. Subsequently, the adjacency
matrix was used to generate face-to-face tie strength for
each day (Total pings/Detected Number of Badges) for
each participant. The badges have been extensively
validated over several studies [7][25][40][37].
KEYS Scale: was used to obtain daily measures of
creativity. The KEYS survey is designed to assess the
perceived stimulants and obstacles to creativity in
organizational work environments. Items of KEYS scale
address negative and positive aspects of the environment. It
is widely recognized as the current standard for measuring
team creativity and innovation within organizational work
environments [4]. The survey has fifteen questions, two of
which are open-ended responses. A 7 point Likert scale is
used for each of the questions. One measure is self-rated
creativity that is extracted from the report of team
creativity. The variable assesses member reports of
creativity being experienced by the team. For the open
ended questions, participants were asked to 1) In a few
words, briefly describe the major work you did on the
assigned project pertaining to this study today, or the major
activities you engaged in that were relevant to the target
project2) Briefly describe ONE event from today that
stands out in your mind as relevant to the target project,
your feelings about this project, your work on this project,
your teams feelings about this project, or your teams work
on this project.”
CODE
EVENTS
CREATI
VE (1/0)
11_P6_D4
Work on Cholecystectomy [sic] scene. Made
a minor breakthrough today. A physics asset
of a gall bladder model seemed to interact
well. I pursued that lead with Cord and as it
turned out I managed to create with his help a
very nicely interacting model of the gall
bladder.
1
12_P6_D4
Besides this I spent the day adding another
layer to the connective tissue.
0
Table 1. A sample of coding combined narratives from events.
CODE value represents Event Number_ParticipantID_Day.
Note: This participant previously described adding the layer
to connective tissue as a routine task.
KEYS requires an expert judge to rate the participants’
reports by assigning numerical value for the level of
creativity (0 or 1) in addition to the self-reports through the
questionnaire. It has been shown to be reliable when the
rating is conducted with one or more expert judges. Expert
is defined as a person knowledgeable about domain.
In this investigation, the method described by Amabile et
al. [4] was followed to obtain “expert coded creativity.”
The descriptions from the open-ended questions were
combined together to form open-ended narratives for each
participant. Unique instances of completed actions were
extracted from the narrative for each participant to identify
individual events.’ The KEYS coding protocol defines
creative thought as any of the following: (1) a discovery,
insight, or idea; (2) the act of searching for a discovery,
insight, or idea; (3) solving a problem in a non-rote way; or
(4) the act of searching for a problem solution in a non-rote
way. Events that had any of these were labeled 1 and events
that did not have any of these were labeled 0. Table 1
illustrates the process through an example.
The basic assumption of KEYS is that psychological
perceptions of work environment by the team members
play a vital role in their creativity. The underlying model
[1] identifies three components within the individual that
have an effect on creativity, including individuals intrinsic
motivation, his or her thinking style, and domain-relevant
knowledge. The KEYS scales have been validated over
several studies and are reported to have high validity and
reliability [2].
EXPLORATORY EXPERIMENT
A pilot study explored whether there are correlations
between individual activity and self-reported team
creativity in a small group. A combination of quantitative
and qualitative data for a team of individuals over an
extended period of time was collected. These individuals
worked in an industry research environment that required
high levels of information technology and creativity in an
industry setting. Creativity was measured through an online
survey that had a combination of scale-rated responses and
open-ended questions that allowed participants to describe
their day-to-day experience of creativity. A multi-
methodological approach was used to explore the
relationship between data obtained from the sensed activity
(movement and face-to-face interaction) [39] and levels of
creativity collected via electronic ESM [4]. The results of
this study informed our hypotheses for subsequent
experiments.
Participants: A team of five people (2 females, 3 males;
mean age = 32.4 years, range= 26-38 years) participated in
a five-week study (total 25 working days). All participants
had undergraduate degrees in engineering and two had
post-graduate degrees (1 MBA and 1 MS). The team was
involved in software coding and research in a leading
industrial research and development laboratory in the
United States. All participants were part of a single team
engaged in highly creative research and development
activities. The participants were selected because they
worked in a tightly knit single location laboratory
advancing IT research. The members conduct the majority
of work in this laboratory in a highly interactive manner.
No rewards were given to participants in this study. The
head of this department was contacted via email and they in
turn put the experimenter in contact with the team that
volunteered for the study. All participation was voluntary
and participants had the option to opt out at any time.
Materials and Procedure: This study used sociometric
badges and the KEYS daily questionnaire. The study was
conducted at a remote site with a team involved
professionally in software coding projects. The
experimenter shipped the badges to the remote site at the
beginning of the study. Participants were required to charge
the badge on their own every night by plugging them into
computers via a USB cable that was provided.
All participants were informed that the investigation was on
workflow issues in teamwork. This was largely done to
avoid any bias on the part of the participants towards
creativity. All participants were given a unique
participation ID through which they corresponded for the
duration of the study. The participants were also informed
that their responses would remain anonymous and
evaluated by researchers unaffiliated with their work
environment. The participants were ensured that the data
would not be shared with the supervisors directly and only
anonymized aggregate analysis would be presented to
audiences.
Before the experiment began, subjects were requested to
answer an initial demographic questionnaire. During the
study, each subject wore a sociometric badge. The subjects
were requested to wear the badges throughout their
workday (9 am to 5 pm) during the experimental period.
At the end of the day, subjects were requested to answer a
daily questionnaire. The data collection protocol occurred
for 25 days and provided us with an extensive sampling of
creative, non-creative episodes and the activity profiles
associated with it. A reminder was sent at 4:15 pm
everyday with the survey link via email to each participant.
The data was downloaded only once at the end of the study
when the badges were shipped back. Except for initial
clarification on how badges worked and debriefing, there
was no interaction between the experimenter and the
participants.
Data Analysis and Results
KEYS and Sociometric Badges data were analyzed.
KEYS Daily Questionnaire Data: While there are 15
questions in the KEYS survey, this analysis focused on
three questions that dealt with self-rated creativity, expert-
coded creativity, and measures of team interaction (other
variables in the KEYS survey are beyond the scope of
current investigation). Out of 125 expected responses (25
days *5 participants), 96 daily surveys were received, and
the average response rate was 76.8% with a standard
deviation of 23%. From the surveys, the value for self
reported creativity was obtained (Likert Scale: 1 – ‘not at
all, 7 extremely). In addition to the scaled responses,
the KEYS instrument and its methodology provide the
opportunity for expert coding of creative and non-creative
events. For each survey response, the narratives from the
two open ended questions were combined. The
participants combined narratives ranged from 462 words
to 2348 words with a mean of 53 words per entry.
Sociometric Badges Data: The five badges were collected
at the end of the 25-day period. While the mean and
standard deviation for four participants was obtained over
all days successfully, one of the badges failed to record any
data and the remaining four failed to record face-to-face
interaction data. Due to the remote nature of the study and
the lack of time stamps, the exact time and duration of
wearing and taking off the badges could not be determined.
Correlation Results: Pearson correlation coefficient values
obtained for all major variables. The significance value was
set at 0.1 as this was an exploratory study with a low N.
There was a significant large correlation (r =0.91) between
self-rated creativity and movement of the participants.
There was a significant medium correlation (r=0.77)
between speech and expert-coded creativity. There was a
significant correlation (r=0.88) between degree (KEYS
scale) and expert-coded creativity. There was a significant
negative correlation (r =- 0.82) between movement and
hours spent with team (KEYS scale).
Summary: Participants self-rated creativity was highly
correlated with their daily movement energy. There was
low correlation between expert-coded creativity and
movement and between expert-coded creativity and self-
reported creativity. The number of people a team member
meets had medium correlation with expert-coded creativity.
The data indicated a few interesting trends. First,
participants feel more creative when they move more.
Second, they are generally more creative when they are
meeting more people in the team and feel more connected.
MAIN EXPERIMENT I
Experiment I employed the pilot study methodologies and
procedures to investigate the following two hypotheses:
H1. Average daily movement energy of team members
during days with above average self-rated creativity is
significantly greater than the average daily movement of
days with below average self-rated creativity
H2. Average face-to-face tie strength of team members
during days with above average expert-coded creativity
is significantly greater than the average face-to-face tie
strength of team members of days with below average
expert-coded creativity
with p < 0.05 accepted as statistically significant.
Methods
Seven participants engaged in creative research were
observed for two work weeks (11 days) during regular
work hours (9 am to 5 pm). The mean age of participants
was 24.7 years (range = 2432 years), and 4 out of 7
participants were men. The sample was highly educated, 4
out of 7 participants were college graduates engaged in
postgraduate work, and 3 were senior undergraduates. All
participants were recruited via email and signed consent
form for voluntary participation prior to the start of the
study. Approval was also obtained from the laboratory
head, prior to the start of the study. No rewards were
provided for participation in the study. Recruited
participants worked together as a team in a Carnegie
Research I University in information technology rich
environments. For 11 days, the experimenter visited the
laboratory each morning to ensure that the participants
badges were worn at 9 am. The experimenter observed
activities throughout the day and at 5 pm requested the
participants to turn the badges off.
Data Analysis and Results
KEYS Daily Questionnaire Data: Out of total 77 (11 days
*7 people) daily online surveys, the number of surveys
completed was 58. The mean response rate was 75% with a
standard deviation of 19%. From the survey, the value of
self-rated creativity was calculated. Expert-coded creativity
scores were obtained by analyzing the narratives.
Participants’ combined narratives ranged from 63 words to
516 words with a mean of 290 words for each participant.
Sociometric Badge Data: For each day, participants’
movement energies were calculated and face-to-face
interaction recorded. The adjacency matrix thus obtained
was made symmetric with respect to the lowest number of
signals (or pings) that were recorded. The average number
of pings was calculated for each participant for each day.
Experimental Data: The following four variables: (1) self-
rated creativity; (2) expert-coded creativity; (3) movement
energy; and (4) face-to-face tie strength were analyzed. The
creativity data was mean split in two ways based on 1) self-
rated creativity and 2) expert-coded creativity. K-Means
clustering showed that the ratio of inter-cluster distance to
intra-cluster distance was high (R=0.94) which validated
the choice of mean split. For each of these two measures of
creativity, the days that had values for creativity higher
than the mean were labeled creative while those days that
were at or below the mean value were classified as non-
creative.
H1 Result: A paired-samples t-test was conducted to test
H1. This t-test [t (36) = 3.132, p < 0.005] confirmed, the
hypothesis that average daily movement energy during
days with above average creativity (M = 1.31, SD = 0.04)
was significantly greater than the average daily movement
of days with below average creativity (M = 1.29, SD =
0.03). The eta-squared statistic (0.21) indicated a large
effect size.
H2 Result: A paired-samples t-test was conducted to test
H2. The t test [t (21) = 1.05, p > 0.1] showed no significant
difference between average face-to-face tie strength of
team members during days with above average expert-
coded creativity (M = 9.4, SD = 10) and the average face-
to-face tie strength of team members (M = 6.3, SD = 7) for
days with below average expert-coded creativity
Correlation Data: There was a significant correlation
between face-to-face interaction and both self-rated
creativity (r=0.45) and expert-coded creativity (r=0.45). A
significant correlation (r=0.66) was also found between
movement and self-rated creativity.
Summary: Results confirm H1, average movement for
creative days is significantly higher than for non-creative
days. H2 was not confirmed. Face-to-face interaction was
highly correlated with expert-coded creativity and
movement was highly correlated with self-rated creativity.
MAIN EXPERIMENT II
Experiment II built on results from Experiment I, using the
same methodologies, procedures, and hypotheses.
Seven participants engaged in creative research were
observed for two work weeks (15 days) during regular
work hours (10 am -5 pm). They engaged in research
intensive creative work in an information technology (IT)
rich environment. The mean age of participants was 27.4
years (range = 2332 years) and 6 out of 7 participants
were men. Our sample was highly educated, 4 out of 7
participants were college graduates engaged in
postgraduate work, and 3 were senior undergraduates. All
participants were recruited via email and signed consent
forms for voluntary participation prior to the start of the
study. Approval was also obtained from the laboratory head
prior to the study and no rewards were provided for
participation.
Materials and Procedure: Experiment II followed
Experiment I protocols, adding ESM reports to understand
face-to-face interaction relationships with creativity. At the
end of each hour (from 11 am - 5 pm), participants received
an SMS request: For the last hour, you were Creative 1 or
Non-creative 2 and Meeting 1 or not meeting 2 (respond 1
1 if creative and meeting and so on)”. Qualitative
observations by an expert coder provided descriptions of
events, people involved, actions, movement, and meetings.
Data Analysis and Results
KEYS Daily Questionnaire Data: Out of a total of 105
(15*7) daily surveys, the number of surveys completed was
76. The open-ended narratives were coded to obtain scores
for expert-coded creativity for all participants for each of
the 15 days. Participantscombined narratives ranged from
293 words to 1663 words with a mean of 973 words.
Sociometric Badge Data: Accelerometer data from the
badge of each participant was downloaded every day. By
using the same formula used in our previous studies, we
calculated a movement energy array that was later used to
give us mean and standard deviation of movement energy
for each day. The average face-to-face tie strength was
calculated for each participant across 15 days.
Experimental Data: We obtained the following four
variables for each participant for 15 days: (1) self-rated
creativity (2) expert-coded creativity (3) movement energy
(4) average face-to-face tie strength. The data was mean
split in two ways based on 1) self-rated creativity and 2)
expert-coded creativity. For each of these two measures of
creativity, the days that had values for creativity higher
than the mean were labeled creative while those days at or
below the mean value were classified as non-creative.
H1 Result. A paired-samples t test was conducted to test
H1. The t test [t (23) = 6.49, p < 0.001] confirmed that
average daily movement energy during days with above
average self-rated creativity (M = 1.37, SD = 0.07) is
significantly greater than the average daily movement of
days with below average self-rated creativity (M = 1.24,
SD = 0.09). This was a large effect (eta-squared = 0.36).
H2 Result. A paired-samples t test was conducted to test
H2. The t test [t (41) = 2.36, p < 0.01] showed average
face-to-face tie strength of team members during days with
above average expert-coded creativity (M = 2.69, SD =
4.01) is significantly greater than the average face-to-face
tie strength of team members for days with below average
expert-coded creativity (M = 0.9, SD = 2.1). The eta-
squared statistic (0.11) indicated a large effect size.
Correlation Data: Pearson product-moment correlations
between all major variables in the study were calculated.
Self-rated creativity was weakly, but significantly
correlated with expert-coded creativity (r = 0.25). In
addition, movement and self-rated creativity were
significantly correlated (r = 0.55). Face-to-face interaction
had significant correlation with both self-rated creativity (r
= 0.20) and expert-coded creativity (r = 0.25).
SMS Data: A total 99 hours of data was collected for 1
hour intervals of self-reported daily activity indicating
whether meetings or non-meetings were occurring and
whether or not these hours were creative. We summed
across all days for four variables: (1) Creative and Meeting
(2) Creative and Not meeting (3) Non-creative and Meeting
and (4) Non-creative and Not meeting. We found that
people reported to be creative while they were meeting
(165 hours) more than twice than when they reported to be
not creative while meeting (71 hours). The correlation
between self-rated creativity and creative and meeting
reports was significant (r = 0.82, p<0.01), and there was
significant negative correlation between reports of non-
creative and non-meeting and self-rated creativity (r= -0.58,
p<0.05).
Summary: A significant difference was found between
team member movements for creative days and non-
creative days. Creative days were also shown to have
higher face-to-face interaction than the non-creative days.
Participants SMS reports indicated that episodes of
meeting one or more team members were twice as likely to
be creative than non-creative. Moreover, across fifteen
days, there was a significant correlation between meeting
episodes and self-reported creativity. People were two
times more likely to report non-creative events when not in
meetings. Overall, results show that participants in a small
group are likely to be more creative when they are more
active, in terms of both movement and face-to-face
interaction.
COMPUTATIONAL MODELING OF TEAM CREATIVITY
Statistical learning techniques and pattern recognition
techniques on a validated subset of features available from
the sociometric badges and labeled events were employed
for the development of the computational models,
following methods developed by Olguin et al. [36][37].
Olguin-Olguin et al. [37] have developed computational
techniques for studying the relation between activity
measured through sociometric badges and several variables
like performance in healthcare environment and studying
productivity in IT domains. These techniques employ
signal processing techniques to extract relevant features
from the data stream that correlate with social signals and
measures of performance. Pentland employed eigenvector
representation to study the variance of behavior of
individuals [39]. They showed that there was limited
amount of behavior variance across days for individuals
implying high predictability.
With data from Experiments I and II, standard linear
regression [14] was used to study relationships between
activity-network profiles and the creativity class.
Subsequently, Naïve Bayesian Classifier (NBC) [14] were
employed. A third approach, a combination of principal
component analysis (PCA) [14] for dimensionality
reduction and linear discriminant analysis (LDA) [14] for
classification based on approaches developed for face
recognition by Li et al. [29], was then used to develop a
deterministic approach to computational modeling of
creativity. Maximum Likelihood Estimation (MLE) was
used as the training approach. Matlab 2009® algorithm
implementations, glmfit for linear regression,
NaiveBayes.fit and predict for NBC and classify function
for LDA, were used.
Procedure for Computational Modeling Approaches
Data from Experiments I and II provided day-to-day team
members’ activity and overall creativity scores. The data
sets were combined and the days were divided into two
classes: creative and non-creative based on the mean split
of reported creativity measures. The movement data was
divided based on self-rated creativity while the face-to-face
interaction (or network pings) data was divided based on
expert-coded creativity. The corresponding measures were
chosen to classify the data based on the results from
Experiments I and II. Overall 182 hours of data per subject
(participant N=7, for 26 days) totaling 1274 hours of data
was collected. The activity as measured through
accelerometer (X, Y, and Z) and network pings from IR
were considered for analysis.
For face-to-face quantification, the average IR ping
information for the day for each participant was considered
for the computational model. As each experiment had 7
participants, there was 7x7 matrix of face-to-face network
pings as sensed by the IR sensor for each day. For every
day, the frequency of pings for every pair of participants,
which could be understood as network edge strength in the
team’s network, was calculated. In the team network,
participants are the node and the edges represent face-to-
face interactions. Of the 26 days worth of network data, 14
were labeled creative and 12 were labeled as non-creative
according to expert-coded creativity. The 7x7 matrix of IR
ping frequency for each day was linearized into a 49x1
vector representing pings per day for all possible person-
person interactions. The creativity class (creative or non-
creative) for each vector was known. This matrix was
employed to train pattern recognition algorithms to assess
expert-coded creativity, measure in Experiment I and II.
For movement analysis, the accelerometer readings were
sampled at 50,000 readings per day with 3 measures per
reading (X,Y,Z), to define a representative sample of
activity profile for each day. The day-activity matrix was
assembled by linearizing the accelerometer activity into a
vector and then assembling the individual vectors into a
matrix. For each vector, a class of creativity (creative or
non-creative) was known. This matrix was employed to
train pattern recognition algorithms to assess creativity.
For each of the three approaches, an 80/20 train/test
paradigm was employed. Algorithms for movement data
and network ping (face-to-face interaction) data were
trained individually as they provided related but distinct
information and related to different measures of creativity.
In the analysis of face-to-face interactions, the core idea
was to classify the entire day as being creative or non-
creative for the whole team. This computational engine was
tuned to gestalt creativity classification for an entire day
and hence provided complementary information to the
movement data based classifier that provided per-day per-
person ratings. The network ping based classifier was
geared towards assessing the team’s everyday creativity.
Results from Computational Modeling Approaches
PCA was applied for dimensionality reduction and LDA for
classification as a means of achieving high accuracy and
fast computation results. The energy of the eigenvalues
indicated 7 dimensions were sufficient to cover 95% of
variance with the movement data; PCA was used to reduce
the dimensionality of the data from 150,000 to 7. LDA was
then used as the classification technique. In the case of
face-to-face interaction strength, 3 dimensions were
sufficient to represent 95% of variance in PCA, hence
dimensionality was reduced to 3 and LDA was performed.
For movement data, linear regression showed the lowest
classification accuracy of 45.2%; the fit was not high or
significant. NBC showed an accuracy of 70.2%. The NBC
fit performed significantly better than linear regression
achieving accuracies of about 70% for both movement and
face-to-face interaction data. The combined approach of
PCA and LDA showed an accuracy of 87.5% for per-day
per-person data as measured through the movement stream
and 90.9% for face-to-face interaction.
Additional endeavors (see [52]) using reduced
dimensionality data for linear regression and NBC showed
slight improvements (47.1% and 71% classification
accuracy). This suggests the superiority of the PCA/LDA
approach in obtaining accurate classification. The results
for face-to-face interaction data were analogous, with the
PCA/LDA combination achieving the highest recognition
accuracy of 90.9%, NBC achieving 71.2% and linear
regression achieving an accuracy of 44.5%.
Discussion of Computational Modeling Approaches
While the results are based on a limited data set and require
further validation, it is encouraging to note that
computational engines can be designed to ascertain
creativity from sensor data. PCA and LDA analysis yielded
close to 92% accuracy. While we cannot imply causality to
this result, the fact that linear approaches give encouraging
results opens up several possibilities for analysis of
creativity in an automated fashion. The two approaches
NBC and PCA-LDA combination both have unique
advantages and requirements. NBC can be very successful
in developing long-term trends and patterns and can be
employed in a formative fashion. The PCA-LDA
combination can actually give the highest accuracy as it
removes noise from the original data. Noise may have
contributed to low accuracy in the results of linear
regression. We trained on data from two studies, suggesting
this approach may be somewhat robust and generalizable.
DISCUSSION
This investigation shows there is a significant relationship
between (1) individual movement and self-rated creativity
and (2) face-to-face interaction and expert-coded creativity.
Specifically, daily movement energy for creative days was
significantly higher than the movement energy of the non-
creative days for team members. In terms of face-to-face
interaction, in Experiment I, while no significant difference
between face-to-face interaction for creative and
noncreative days classified on the basis of expert-coded
creativity was found, the trend in significance encouraged
further exploration in Experiment II, using sms based ESM
to gather participants report on their ongoing behavior [54].
Participants reported the highest number of both creative
and meeting and noncreative and not-meeting episodes.
The third highest number was creative and not-meeting.
The least reported variable was not creative and meeting.
This suggests that participants were generally more creative
when they were meeting and generally more non-creative
when they were not meeting. However, an important
component of overall team creativity is a combination of
team and individual creativity. Reports of being creative
and not-meeting tend to be on the same days participants
had also reported to be creative and meeting. Teams were
far more creative on days in which the members met. Team
members also reported to be personally more creative after
active interactions with other team members. In Experiment
II, face-to-face tie strength was significantly greater in the
creative days than that of non-creative days.
These results show a strong correlation between movement
and self-rated creativity. Prior studies have found that
exercise or a physical activity of some kind enhances
cognitive performance [49] and physical activity is
correlated with creativity [30]. While the results of these
prior studies were based on questionnaire data implemented
on the middle aged or elderly populations, the presented
research is the first effort of its kind to employ a multi-
methodological approach that confirms the relationship
between sensed movement data with creativity.
In the KEYS scale, there are two variables namely
individual creativity and team creativity. Interestingly,
there was 94% correlation between reported scores of team
creativity and individual creativity. This might be because
individual participants had no prior definition of creativity
or basis of differentiating between the two variables. This
raises the question, “What, exactly, do the personal
creativity and team creativity variables represent in the
KEYS scale?” The results show that how participants feel
about their own creativity (self-reported personal creativity)
may be the same as how they rate their team as being
creative (team creativity). It must be noted that the
creativity score obtained by an expert that is based on their
descriptions is not necessarily correlated with their
creativity self-reports. This could be because these measure
two different facets of creativity.
Several key questions need to be answered with respect to
the nature of the following question: how much and how
often team interactions should occur for the team to be
more creative? To provide clues to some of these questions,
the role of degree (number of team members meeting each
other) in both the experiments was explored and found in
both cases to have no significant [t (36) = 1.55, p > 0.12; t
(20) = 0.23, p > 0.8]. The means in the two cases in both
the experiments were almost equivalent (Non Creative: M
= 1.81, SD = 1.68; M = 1.67, SD = 1.8; Creative: M =
1.24, SD = 1.38; M = 1.52, SD = 1.5). Thus it is not the
number of people a team member meets with, but rather the
quality of face-to-face interaction (or the time spent with
the team members) that influences creativity.
The computational modeling results show the feasibility of
developing an automated system for creativity based on
team members’ activities. Linear regression analysis did
not yield high accuracy but the Bayesian modeling and the
combined PCA and LDA approach had high recognition
accuracies. Principal component analysis showed that 7
dimensions encompassed close to 95% variation in the
underlying data from movement data and 3 dimensions
encompassed 95% variation in the face-to-face interaction
data. The analysis of principal dimensions and the weights
of the individual units showed that individuals with highest
creativity were given the highest weight and individuals
with the lowest creativity where given the lowest weight.
CONCLUSION
Computational modeling of team creativity has several
benefits: (1) it allows automated evaluation and prediction
of creativity; (2) it paves the way for software and
programs that support creativity; (3) it allows development
of guidelines and procedures towards creativity in teams
and within organizations; (4) it enhances theoretical
understandings of creativity and its relationship to team
member activity.
The key findings of these studies were: (1) days in which
the team is highly creative are also the days in which the
teams’ members meet more often, and (2) days in which
team members report to be highly creative have higher
levels of movement among team members than the days
they report to be non-creative. Through a multi-
methodological approach that coupled sensor based
analysis, wearable computing, and creative behavior
assessment this investigation helps us better understand the
nature and mechanisms of team creativity in the wild.
ACKNOWLEDGMENTS
We thank Teresa Amabile, Leo Burd, Mary Czerwinski,
Kanav Kahol, Taemie Kim, and Sandy Pentland for advice
and support in conducting this research. This material is
based upon work supported by the National Science
Foundation under Grant No. 0846148.
REFERENCES
1. Amabile, T. (1993). Motivational synergy: toward
new conceptualizations of intrinsic and extrinsic
motivation in the workplace. Human Resource
Management Review, 3(3), 185-201.
2. Amabile, T. (1996). Creativity in context: Westview
Press, Boulder, CO.
3. Amabile, T. M. (1983). The social psychology of
creativity: a componential conceptualization. Journal
of Personality and Social Psychology, 45(2), 357-376.
4. Amabile, T. M., Barsade, S. G., Mueller, J. S., &
Staw, B. M. (2005). Affect and creativity at work.
Administrative Science Quarterly, 50, 367-403.
5. Amabile, T. M., Conti, R., Coon, H., Lazenby, J., &
Herron, M. (1996). Assessing the work environment
for creativity. Acad. of Mgt. J., 39(5), 1154-1184.
6. Bain, P., Mann, L., & Pirola-Merlo, A. (2001). The
innovation imperative. Small Group Research, 32(1),
5573.
7. Basu, A. 2002. Conversational scene analysis.
Doctoral dissertation. MIT.
8. Besemer, S. P., & O’Quin, K. (1993). Assessing
creative products: Progress and potentials. In S. G.
Isaksen (Ed.), Nurturing and Developing Creativity:
The Emergence of a Discipline (pp. 331-349).
Norwood, New Jersey: Ablex Publishing Corp.
9. Burleson, W., Picard, R., Perlin, K., & Lippincott, J.
(2004). A platform for affective agent research.
Wkshp. on Empathetic Agents, 3rd. Int. Joint Conf. on
Autonomous Agents and Multi-Agent Sys.
10. Burt, R. S. (2004). Structural holes and good ideas.
American Journal of Sociology, 110(2), 349-399.
11. Choudhary, T. 2004. Sensing and modeling human
networks. Doctoral dissertation. MIT.
12. Christensen, R., & Bedrick, E. J. (1997). Testing the
independence assumption in linear models. Journal of
the American Statistical Assoc., 92, 1006-1016.
13. DiMicco, J. M., & Bender, W. (2007). Group
reactions to visual feedback tools. In Y. de Kort, B. J.
Fogg, W. IJsselsteijn, B. Eggen & C. Midden (Eds.),
Persuasive Technology (Vol. 4744, pp. 132-143):
Springer Berlin / Heidelberg.
14. Duda, R., Hart , P., & Stork, D. (2000). Pattern
Classification (2 ed.). NY: Wiley Interscience.
15. Eagle, N. (2005). Machine Perception and Learning
of Complex Social Systems. MIT.
16. Farooq, U., Carrol, J. M., Ganoe, C. H. 2007.
Supporting creativity with awareness in distributed
collaboration. ACM Group.
17. Granovetter, M. (1973). The strength of weak ties.
American Journal of Sociology, 78(6), 1360-1380.
18. Griffin, D., & Gonzalez, R. (1995). Correlational
analysis of dyad-level data in the exchangeable case.
Psychological Bulletin, 118, 430-439.
19. Guilford, J. P. (1956). The structure of intellect.
Psychological Bulletin, 53(4), 267-293.
20. Hoyle, R., Georgesen, J., & Webster, J. (2001).
Analyzing data from individuals in groups: The past,
the present, and the future. Group Dynamics: Theory,
Research, and Practice, 5(1), 41-47.
21. Kahneman, D., Krueger, A., Schkade, D., Schwarz,
N., Stone, A. (2004). A survey method for
characterizing daily life experience: the day
reconstruction method science, 306 (5702), 1776-80.
22. Kapoor, A., Burleson, W., & Picard, R. W. (2007).
Automatic prediction of frustration. International
Journal of Human-Computer Studies, 65, 724-736.
23. Kenny, D. A., & Judd, C. M. (1996). A general
procedure for the estimation of interdependence.
Psychological Bulletin, 119, 138-148.
24. Kenny, D. A., Mannetti, L., Pierro, A., Livi, S., &
Kashy, D. A. (2002). The statistical analysis of data
from small groups. Journal of Personality and Social
Psychology, 83(1), 126137.
25. Kim, T., Chang, A., & Pentland, A. S. (2007).
Enhancing organizational communication using
sociometric badges. Paper presented at the IEEE 11th
Int. Symp. on Wearable Computing.
26. Kraut, R. E., Egido, C., & Galegher, J. (1988).
Patterns of contact and communication in scientific
research collaboration. CSCW.
27. Lee, L., & Grimson, W. E. (2002). Gait analysis for
recognition and classification. 5th IEEE Int. Conf. on
Automatic Face and Gesture Recognition (FG'02).
28. Leshed, G. (2009). Automated language-based
feedback for teamwork behaviors. Doctoral
dissertation. Cornell University.
29. Li, J., Zhao, B., & Zhang, H. (2009). Face recognition
based on PCA and LDA combination feature
extraction. In 1st. Int. Conf. on Inf. Sci. and Eng.
30. Malone, H. N. (1989). The relationship between traits
of creativity and physical activity in the elderly.
(Unpublished thesis). Ohio State University.
31. Massetti, B. (1996). An empirical examination of the
value of creativity support systems on idea
generation. MIS Quarterly, 20, 83-97.
32. Mayer, R. E. (1999). Fifty years of creativity
research. In R. J. Sternberg (Ed.), Handbook of
Creativity. NY: Cambridge University Press.
33. Moritz, S. E., & Watson, C. B. (1998). Levels of
analysis issues in group psychology. Group Dyn.:
Theory, Research, and Practice, 2, 285-298.
34. Obstfeld, D. (2005). Social networks, the tertius
iungens orientation, and involvement in innovation.
Administrative Science Quarterly, 50, 100-130.
35. Ohly, S., Kase, R., & Skerlavaj, M. (2010). Networks
for generating and for validating ideas: The social
side of creativity. Innovation: Management, Policy &
Practice, 12, 41-52.
36. Olguín-Olguín, D., Kam, M., & Pentland, A. S.
(2010). Quantifying the effects of centrality and tie
strength on performance in face-to-face networks.
Workshop on Information Networks, New York, NY.
37. Olguin-Olguin, D., Gloor, P. A., & Pentland, A. S.
(2009). Capturing individual and group behavior with
wearable sensors. AAAI Spring Symposium.
38. Pentland, A. (2005). Socially aware computation and
communication. IEEE Computer (March), 63-70.
39. Pentland, A. (2007). Automatic mapping and
modeling of human networks. Physica A: Statistical
Mechanics and Its Applications, 378(1), 59-67.
40. Pentland, A. (2008). Honest Signals. Cambridge,
Massachusetts: The MIT Press.
41. Pentland, A. (2009). Reality mining of mobile
communications: Toward a new deal on data. The
Global Information Technology Report, 75-80.
42. Perry-Smith, J. E., & Shalley, C. E. (2003). The social
side of creativity: A static and dynamic social
network perspective. Academy of Management
Review, 28, 89-106.
43. Pirola-Merlo, A., & Mann, L. (2004). The
relationship between individual creativity and team
creativity: aggregating across people and time.
Journal of Organizational Behavior 25, 235257.
44. Ravi, N., Dandekar, N., Mysore, P., & Littman, M. L.
(2005). Activity recognition from accelerometer data.
In The Seventeenth Conference on Innovative
Applications of Artificial Intelligence (IAAI),
Pittsburg, Pennsylvania, July 2005.
45. Saddler, M. S., & Judd, C. M. (2003). Overcoming
dependent data: A guide to the analysis of group data.
In M. A. Hogg & R. S. Tindale (Eds.), Blackwell
Handbook of Social Psychology: Group Processes.
Oxford, UK: Blackwell.
46. Scott, S. G., & Bruce, R. A. (1994). Determinants of
innovative behavior: a path model of individual
innovation in the workplace. Academy of
Management Journal, 37(3), 580-607.
47. Shneiderman, B. (2003). Leonardo's Laptop: Human
Needs and the New Computing Technologies.
Cambridge, MA: The MIT Press.
48. Shneiderman, B., Hewett, T., Fischer, G., Jennings,
P., Czerwinski, M., Kules, B., et al. (2006). Creativity
support tools, NSF wkshp. Int. J. HCI, 20(2), 61-77.
49. Singh-Manoux, A., Hillsdon, M., Brunner, E., &
Marmot, M. (2005). Effects of physical activity on
cognitive functioning in middle age: Evidence from
the Whitehall II Prospective Cohort Study. American
Journal of Public Health, 95(12), 2252.
50. Taggar, S. (2002). Individual creativity and group
ability to utilize individual creative resources: a
multilevel model. Acad. of Mgt. J., 45(2), 315330.
51. Torrance, E. P. (1974). The Torrance Tests of
Creative Thinking-Norms-Technical Manual
Research Edition-Verbal Tests, Forms A and B-
Figural Tests, Forms A and B. Princeton, NJ:
Personnel Press.
52. Tripathi, P. (2011). Predicting creativity in the wild:
Doctoral dissertation. ASU.
53. Turk, M., & Pentland, A. (1991). Eigenfaces for
recognition. J. of Cog. Neurosicence, 3(1), 71-86.
54. Weisner, R., Ryan, G., Reese, L., Kroesen, K.,
Bernheimer, L., & Gallimore, R. (2001). Behavior
sampling and ethnogr. Field Methods, 13(1), 20-46.
55. Zhou, J., Shin, S., Brass, D., Choi, J., & Zhang, Z.
(2009). Social networks, personal values, and
creativity. J. of Applied Psych., 94(6), 1544-1552.
... The advent and improvement of new technologies could promote the collection of precise, time-varying data on team behaviors, especially on teams "in the wild". For example, Olguín et al. (Olguín et al. 2009) have begun developing and using new technologies, such as wearable badges that record audio, movement and proximity data, to record behaviors and interactions between individuals (Tripathi and Burleson 2012;Curhan and Pentland 2007;Anmol et al. 2004;Woolley et al. 2010). In addition to wearable badges, data can be collected during observations using hand-held devices and data-coding software (Jones et al. 2018;Tejani et al. 2010), through video recordings (Jayagopi et al. 2009), or extracted from email interactions or collaborative documents (Wise 2014). ...
... Many studies using new tools to obtain more objective and continuous levels of data on team behaviors and interactions are lacking in their theoretical framework. For example, Tripathi and Burleson (Tripathi and Burleson 2012) were able to predict days of high and low creativity (determined from self-rated and expert-coded creativity scores) from how much physical movement and face-to-face interaction time three teams displayed within an industrial research and development laboratory. While such measurements may provide more accurate measures of the team members' levels of movement and interaction, compared to self-reports or observations, this level of understanding does not provide insight into the mechanisms through which movement or face-toface interactions impact creativity. ...
Article
Full-text available
Researchers of team behavior have long been interested in the essential components of effective teamwork. Much existing research focuses on examining correlations between team member traits, team processes, and team outcomes, such as collective intelligence or team performance. However, these approaches are insufficient for providing insight into the dynamic, causal mechanisms through which the components of teamwork interact with one another and impact the emergence of team outcomes. Advances in the field of animal behavior have enabled a precise understanding of the behavioral mechanisms that enable groups to perform feats that surpass the capabilities of the individuals that comprise them. In this manuscript, we highlight how studies of animal swarm intelligence can inform research on collective intelligence in human teams. By improving the ability to obtain precise, time-varying measurements of team behaviors and outcomes and building upon approaches used in studies of swarm intelligence to analyze and model individual and group-level behaviors, researchers can gain insight into the mechanisms underlying the emergence of collective intelligence. Such understanding could inspire targeted interventions to improve team effectiveness and support the development of a comparative framework of group-level intelligence in animal and human groups.
... Trust occurs more with face-to-face interactions and physical co-location and in turn promotes knowledge sharing and transfer among employees (Golden and Raghuram, 2010). Co-location of employees increases creativity and innovation (Schmidt and Rosenberg, 2014;Tripathi and Burleson, 2012;Waber, 2013). With reference to these previous studies, we choose two variables namely "Lack of Social Interaction" (LOSI) and "Barrier to Effectiveness of Work from Home"(BEWH) and hypothesize that physical separation leads to a lack of social interaction in work from home and is detrimental to employee work effectiveness. ...
Article
Purpose The COVID-19 pandemic pushed organizations across the world to suddenly adopt work from home at a mass scale to maintain business continuity. This study aims to investigate the influence of lack of social interaction in work from the home arrangement on employee work effectiveness. Design/methodology/approach The data were analysed using the partial least square (PLS) structural equation modelling (SEM) approach, a nonparametric method based on total variance, through SmartPLS software 3.0. The data were collected from 399 participants in India using the snowball sampling technique. The target populations were the people who were working from home due to the pandemic and by asking them to forward the survey link in their network. Findings The results suggest that social interaction has a significant positive impact on work effectiveness. However, this impact is not affected by the employee's perceived benefits of maintaining social distancing during the pandemic. The study also found that social interaction is important for both genders and found no significant difference in the relationship between social interaction and work from home effectiveness for male and female employees. Practical implications This study will be useful for human resource practitioners and managers as they build strategies to adopt work from home as a regular practice even in non-pandemic situations. Originality/value This study is the first to examine the influence of lack of social interaction on the work from home effectiveness during the COVID-19 times. It examines the moderating role of the perceived benefits of maintaining social distancing and gender on the effect of lack of social interaction as a barrier to work from home effectiveness.
... It has been argued that in researching collaboration, it is crucial to zoom in to the interactions in order to understand collaboration [9]. In regards of quantifying and measuring collaboration patterns in face-to-face interactions, extensive research has been conducted with the use of sociometric badges, with a number of important findings about dynamics in small collaborative groups [30,31,50]. Other common approaches include social network analysis, which can be used to trace and visualize networking interaction taking place in collaborative teamwork (e.g. ...
Chapter
This paper presents our systematic review of empirical learning analytic studies carried out at K-12 education with a specific focus on pedagogically innovative (constructive) approaches on technology-mediated learning, such as knowledge building, knowledge creation, and maker-centered learning and maker culture. After reading abstracts of identified 236 articles, we zoomed in on 22 articles. We identified three categories of studies: 1) articles oriented toward methodology development, 2) articles relying on digital tools (learning environments with LA functions) and 3) articles investigating the impact of LA.
... Sociometric badges are well established in the literature, having been successfully deployed in a variety of organizational contexts with a variety of predictor variables. The badges have been used to predict organizationally relevant outcomes such as job attitude and performance [25], job satisfaction [24], network cohesion [40], creativity [38], group performance [23], and group collaboration [16]. ...
Conference Paper
Full-text available
Collaborative creativity is an essential part of modern teamwork and is often supported by formal techniques, such as design thinking. Current support tools are often limited in scope as understanding the time-varying nature and structure of team communication is insufficient. We investigate how collaborative creative activities in new product development teams can be digitally supported while maintaining face-to-face communication. This work analyzes to what extent paralinguistic and proxemic features of team interaction relate to performance in new product development teams and if and how this relationship differs for different stages in the design process. This is investigated by applying multilevel modeling on data collected during a four-week new product development cycle. The cycle was completed by four teams, during which data were collected automatically using sociometric badges that capture social signals of team interactions. In addition, the data are combined with survey-based measurements on the team’s daily design process and periodic performance evaluations. The current paper provides evidence that social signals are related to team performance and that this relationship varies across the stages in the product design process. Certain social signals contribute positively in one stage but less in other stages, showing the importance of using multimodal signals when modeling high-level collaborative patterns. This research contributes to the literature by providing a better understanding of relevant factors when designing supporting tools or methods for collaborative creative problem solving.
... We are collecting these quantitative, non-verbal features using a wearable sensor called the sociometric badge that has been successfully deployed in a variety of organizational contexts with a variety of predictor variables. The badges have been used to predict organizationally relevant outcomes, such as job attitudes and performance [27], job satisfaction [26], network cohesion [40], creativity [39], group performance [25], and group collaboration [18]. This study introduces the novel context of predicting DT as a context for the sociometric badge. ...
Chapter
Full-text available
Despite the importance of team communication for successful collaborative problem solving, automated solutions for teams are notably absent from the literature. One promising avenue of research has been the development and integration of speech-based technology for team meetings. However, these technologies often fall short of meeting the needs of the teams as they do not take meeting context into consideration. In this paper, we demonstrate the efficacy of context detection with data collected during real team meetings. By capturing and analyzing social signals of rotation in team dynamics, we can demonstrate that different stages of collaborative problem solving using the design thinking methodology differ in their dynamics. Using supervised machine learning, we successfully predict design thinking mode with an overall F1 score of 0.68 and a best-performing sub-class model of 0.94. We believe this to be an essential step towards improving speech-based technology that aims to assist teams during meetings. Making these automated systems context-aware will enable them to provide teams with relevant information, such as resources or guidance. KeywordsSocial signalsContext detectionPredictive modeling
Article
Smart speakers have become exceedingly popular and entered many people's homes due to their ability to engage users with natural conversations. Researchers have also looked into using smart speakers as an interface to collect self-reported health data through conversations. Responding to surveys prompted by smart speakers requires users to listen to questions and answer in voice without any visual stimuli. Compared to traditional web-based surveys, where users can see questions and answers visually, voice surveys may be more cognitively challenging. Therefore, to collect reliable survey data, it is important to understand what types of questions are suitable to be administered by smart speakers. We selected five common survey questionnaires and deployed them as voice surveys and web surveys in a within-subject study. Our 24 participants answered questions using voice and web questionnaires in one session. They then repeated the same study session after 1 week to provide a "retest'' response. Our results suggest that voice surveys have comparable reliability to web surveys. We find that, when using 5-point or 7-point scales, voice surveys take about twice as long as web surveys. Based on objective measurements, such as response agreement and test-retest reliability, and subjective evaluations of user experience, we recommend that researchers consider adopting the binary scale and 5-point numerical scales for voice surveys on smart speakers.
Book
Full-text available
This book provides insight for researchers and decision-makers on the application of data in the entrepreneurship and sustainable development sector. This book covers how Big Data for Industry 4.0 and entrepreneurship are effective in resolving business, social, and economic problems. The book discusses how entrepreneurs use Big Data to cut costs and minimize the waste of time. It offers how using Big Data can increase efficiency, enables the studying of competitors, can improve the pricing of products, increase sales and loyalty, and can ensure the right people are hired. The book presents how decision-makers can make use of Big Data to resolve economic and social problems. Analyze the development of the economy and enhance the business climate. This book is for researchers, PhD students, and entrepreneurs and can also be of interest for transforming governments as well as businesses.
Article
Full-text available
Evolutionary models and empirical evidence suggest that outgroup threat is one of the strongest factors inducing group cohesion; however, little is known about the process of forming such cohesive groups. We investigated how outgroup threat galvanizes individuals to affiliate with others to form engaged units that are willing to act on behalf of their in-group. A total of 864 participants from six countries were randomly assigned to an outgroup threat, environmental threat, or no-threat condition. We measured the process of group formation through physical proximity and movement mirroring along with activity toward threat resolution, and found that outgroup threat induced activity and heightened mirroring in males. We also observed higher mirroring and proximity in participants who perceived the outgroup threat as a real danger, albeit the latter results were imprecisely estimated. Together, these findings help understand how sharing subtle behavioral cues influences collaborative aggregation of people under threat.
Article
2019年12月に中国で発生した新型コロナウイルス感染症は全世界に蔓延し、日本では2020年4 月7日に緊急事態宣言が発令された。日本政府からの自粛要請により外出自粛が求められた結果、企業は社員を通勤させずに業務を続けられるよう、全国でテレワークの導入が行われ、一橋大学とHR総研の調査によれば特に首都圏で急増した。元来、日本政府は2020年東京オリンピック・パ ラリンピック競技大会時における交通混雑緩和を目指してテレワーク関連府省連絡会議を開催し、テレワークを推進していたが、その普及割合は総務省の調査によれば19.1%と低いものであった。一橋大学とHR総研の調査が緊急事態宣言発令直後の2020年4月17日~24日かけて実施した314社 の人事部門に対するアンケート調査による研究では、新型コロナウイルス感染症の影響を受け84% の企業が在宅勤務やテレワークを導入している。感染症と外出自粛という予期せぬ外部環境の変化 によって、事業を継続させるために日本企業の中で急激にテレワークが広がったことがうかがえる。 テレワークに関連する研究は Web of Science で ”Telecommuting”と検索すると635の論文が、”Telework”と検索すると667の論文が該当し(2020年8月4日検索)、その領域も経営学だけでなく社会学、心理学、交通や環境等多岐にわたる。これまでの研究では従業員の満足度に焦点を当てた研究が多く、その結果もプラス・マイナスの影響が混じっている。従業員の満足度は企業のイノベーションにプラスの影響を与えるものであるが、テレワークでは対面でのコミュニケーションが取れず、知識移転が容易ではなくなり、イノベーションへの影響が懸念される。経済産業省の日本企業における価値創造マネジメントに関する行動指針では、イノベーションを社会・顧客の課題 解決につながる革新的な手法(技術・アイデア)で新たな価値(製品・サービス)を創造すると定義しており、本稿ではテレワークと従業員の職務満足度、イノベーションの関係性に焦点を当てた。 企業の持続的な発展にはイノベーションが欠かせず、今後急増することが見込まれるテレワークという新しい働き方の中でいかにイノベーションを生み出すかは重要である。テレワークは従業員の職務満足度を向上させ、結果としてイノベーションにプラスの影響があることが示唆されている。 他方、職務満足度が上がると「現状で問題ない」という現状維持を促し、その結果としてサーチを怠り(入山, 2019)、イノベーションに負の影響を与える可能性がある。また、対面コミュニケーショ ンの減少から、ウェブミーティングやメールなどを使ったICTによるコミュニケーションであっても伝えられる情報は言語や画面を通したものに限られ情報交換が難しくなり、イノベーションに負の影響を与えることも示されている。テレワークの研究はいまだ発展途上であり、職務満足度や、テレワークの頻度がイノベーションへどういった影響を与えるかもいまだ明らかにされていない。
Article
Engineering design typically occurs as a collaborative process situated in specific context such as computer-supported environments, however there is limited research examining the dynamics of design collaboration in specific contexts. In this study, drawing from situative learning theory, we developed two analytic lenses to broaden theoretical insights into collaborative design practices in computer-supported environments: (a) the role of spatial and material context, and (b) the role of social interactions. We randomly assigned participants to four conditions varying the material context (paper vs. tablet sketching tools) and spatial environment (private room vs commons area) as they worked collaboratively to generate ideas for a toy design task. We used wearable sociometric badges to automatically and unobtrusively collect social interaction data. Using partial least squares regression, we generated two predictive models for collaboration quality and creative fluency. We found that context matters materially to perceptions of collaboration, where those using collaboration-support tools perceived higher quality collaboration. But context matters spatially to creativity, and those situated in private spaces are more fluent in generating ideas than those in commons areas. We also found that interaction dynamics differ: synchronous interaction is important to quality collaboration, but reciprocal interaction is important to creative fluency. These findings provide important insights into the processual factors in collaborative design in computer-supported environments, and the predictive role of context and conversation dynamics. We discuss the theoretical contributions to computer-supported collaborative design, the methodological contributions of wearable sensor tools, and the practical contributions to structuring computer-supported environments for engineering design practice.
Article
Full-text available
The authors elaborate the complications and the opportunities inherent in the statistical analysis of small-group data. They begin by discussing nonindependence of group members’ scores and then consider standard methods for the analysis of small-group data and determine that these methods do not take into account this nonindependence. A new method is proposed that uses multilevel modeling and allows for negative nonindependence and mutual influence. Finally, the complications of interactions, different group sizes, and differential effects are considered. The authors strongly urge that the analysis model of data from small-group studies should mirror the psychological processes that generate those data.
Article
The present study integrated a number of streams of research on the antecedents of innovation to develop and test a model of individual innovative behavior. Hypothesizing that leadership, individual problem-solving style, and work group relations affect innovative behavior directly and indirectly through their influence on perceptions of the climate for innovation, we used structural equation analysis to test the parameters of the proposed model simultaneously and also explored the moderating effect of task characteristics. The model explained approximately 37 percent of the variance in innovative behavior. Tasktype moderated the relationship between leader role expectations and innovative behavior.
Conference Paper
Humanity has the beginnings of a new nervous system - a digital one derived from mobile telephone networks and already nearly four billion people strong. Computational models based on these digital "people data," using a process called reality mining, allow us to create a startlingly comprehensive picture of our lives, predicting human social behavior with a power that was barely conceivable just a few years ago. This new "god's eye" view of humanity will present unprecedented opportunities for profitable commerce and effective government ... and it will endanger our personal freedom [1]. To harness the good and avoid the evil, we need to forge a 'New Deal' about how personal data can be gathered and used [2].
Article
Introduction Several researchers have studied the relationship between social network attributes (both at the individual and group levels), and performance. Traditional social network data collection techniques include questionnaires, interviews, web sites, and e-mail repositories. However, the study of large face-to-face social networks has only recently become possible with the advent of pervasive computing and portable sensing devices like the MIT sociometric badges (Olguin-Olguin et al., 2009). One of the basic elements of cooperation in groups is face-to-face interaction. Individuals promote each other's success by helping and encouraging each others' efforts. It is well known that accountability to peers, the ability to influence each other's reasoning, social modeling, social support, and interpersonal rewards, all increase as the face-to-face interactions among the group members increase (Johnson & Johnson, 1999). Nevertheless, too much face-to-face interaction might lead to information overload and have negative effects on individual performance (Eppler & Mengis, 2004). This study is an attempt to uncover some of the important social dynamics mechanisms driving an individual's performance by looking at individual centrality and face-to-face interaction strength in the context of distributed knowledge (i.e. transactive memory systems) and cognitive load theory.
Article
We propose using an existing set of statistical tools in a new way that allows one to test the independence assumption in standard normal theory linear models. The set of tools is near-replicate lack-of-fit tests. The classical lack-of-fit test requires a linear model in which some rows of the model matrix are repeated. Near-replicate lack-of-fit tests were developed to mimic the behavior of the classical test by identifying clusters of rows in the design matrix that are similar, though not necessarily exact replications. We argue that meaningful clusters can be formed more generally by constructing rational subgroups of data collected under similar circumstances. As such, observations in the same subgroup may be more highly correlated than observations in different subgroups. We investigate the behavior of these tests when used to identify lack of independence.
Article
We have developed a near-real-time computer system that can locate and track a subject's head, and then recognize the person by comparing characteristics of the face to those of known individuals. The computational approach taken in this system is motivated by both physiology and information theory, as well as by the practical requirements of near-real-time performance and accuracy. Our approach treats the face recognition problem as an intrinsically two-dimensional (2-D) recognition problem rather than requiring recovery of three-dimensional geometry, taking advantage of the fact that faces are normally upright and thus may be described by a small set of 2-D characteristic views. The system functions by projecting face images onto a feature space that spans the significant variations among known face images. The significant features are known as "eigenfaces," because they are the eigenvectors (principal components) of the set of faces; they do not necessarily correspond to features such as eyes, ears, and noses. The projection operation characterizes an individual face by a weighted sum of the eigenface features, and so to recognize a particular face it is necessary only to compare these weights to those of known individuals. Some particular advantages of our approach are that it provides for the ability to learn and later recognize new faces in an unsupervised manner, and that it is easy to implement using a neural network architecture.
Article
An improved healthcare system will require significant departures from current practices. Leadership capacity will provide the abilities to innovate and creative thinking needed to be effective in championing the value of occupational therapy in the future.