A collaboration is a collaboration is a collaboration1
1. “A collaboration is a collaboration is a collaboration. Not.”
John C. Thomas
IBM T. J. Watson Research Center
Yorktown Heights, New York 10598
jcthomas@us.ibm.com
www.truthtable.com
Abstract: Human-Computer Interaction as a field has migrated over the last few decades
from being primarily concerned with individual users interacting with dedicated
computers to a field that often examines and tries to optimize teams, groups, communities
and society interacting via a wide variety of devices in a wide variety of settings. Better
collaboration is variously perceived as necessary for business, an inevitable aspect of the
new economy and even crucial for our survival as a species. These propositions are all
probably true but they obscure another truth; viz., that successful “collaboration” entails
not just one, but a wide variety of skills and attitudes. In this paper, I argue that games
meant to use or teach “collaboration” would be more successful if more differentiated
and specific about the aspects of collaboration they are focusing on. I suggest a
categorization scheme for varieties of collaboration and related skills and attitudes.
Finally, I briefly outline some plans for examining the effects of team-building games in
Second Life.
Why games? Why have virtually all societies developed games? Succinctly put, games
enable learners to concentrate on building skills by providing quicker and less ambiguous
feedback than is typically possible in real life. The skills thus learned may then be
applied to real life situations. A quintessential and well-known example is chess which
embodies principles of threat, double attack, pinning, the importance of position, superior
force, the importance of mobility and timing. All of these are important in real warfare
(or economic competition) as well, but in real war, any particular commander is likely to
have only a very limited number of experiences while anyone can play hundreds of chess
games. Moreover, in real war, the actual outcome often depends on other non-strategic
factors such as weather, luck, disease or one side beginning with an overwhelming
resource advantage. Therefore chess, like many other games, has the advantage of
showing the interplay of important principles quickly, often, and unambiguously.
In order to accomplish this, chess is an extremely “pared down” version of real war. For
example, the two armies begin with only 16 soldiers on each side. The space is limited to
64 discrete squares. The two opponents are the only “conscious” entities. Each “soldier”
does exactly as it is told 100% of the time. Time is partitioned into discrete moves. The
board is completely visible at all times. There is no “fog of war.” The so-called “moral”
aspects of war (that is, the psychological aspects) are not really present (U.S. Marine
Corps, 1994).
By contrast, modern technology allows us to forgo many of these simplifications and
provide games, such as “World of Warcraft” which are much more like “real” war. The
2. terrain is huge and varied; there are scores of players interacting continuously;
communication and misdirection become vital components and so on. Does this
necessarily imply that “World of Warcraft” is a better training ground than chess for real
military thinking because it covers more factors? That is an empirical and (so far as I
know), unanswered question. However, one thing “World of Warcraft” certainly does
allow is collaboration. In chess, there is no collaboration except insofar as the player
forces/enables his soldiers to cooperate effectively. In order to examine such potentially
empirical questions more deeply, however, we must first unpack the notion of
“collaboration” and see what is involved.
The components of collaboration. First, it seems that effective collaboration in the real
world depends on many factors beyond the skills of the people involved. It depends on
external motivational structures, tools and technology. Some of the skills involved in
successfully collaborating in the real world involve those factors. For instance, a person
may be a “good collaborator” in some sense, but if they cannot use the tools that their
company has provided for communication, their impact may be quite limited. This is
potentially more than a trivial point. If, for instance, 90% of a person’s time is in
learning the interface of “World of Warcraft” or “Second Life,” there is correspondingly
less time to learn transferable collaboration skills per se.
If we focus only on “focal” collaborative skills however, there are still at least two
important major aspects; action skills and attitudes. In the realm of attitudes, there is
trust and this itself has often been further broken down into such factors as attitudinal or
personality trust (some people are more trusting in general), competency trust (I trust that
you are able to do something), personal trust (I trust you to do the “right thing” regardless
of consequences) and institutional trust (I trust you to do the right thing because there are
safeguards in place such as rewards and punishments to help insure that it is in your own
interest to do the right thing). See Riegelsberger & Vasilou, (2007) and Riegelsberger,
Sasse & McCarthy, (In Press) for detailed models of trust.
In the realm of skills, one way to break down collaboration skills is to consider the ability
to perceive the actions of others; the ability to infer their intentions and likely future
actions; the ability to signal one’s own actions and intentions; the ability to change levels
or frameworks; that is, to recognize that something systematic is going wrong and to
generate and implement fixes. These skills may also be broken down by time scale; for
instance, in the arena of long-term strategy, shorter term tactics, and moment to moment
on-going action. Individuals may also have various levels of skill (due to training or
predisposition or both) in terms of modality. For example, jazz musicians in a jam may
be quite good at picking up on what their collaborators are doing based on hearing
subtleties while soccer players may be relying on subtleties of motion provided by vision;
in bobsled racing, kinesthesia may be relatively more important and so on.
The realm of games provides many different models for the kinds of collaboration that
are particularly important. For instance, most of the events in track and field are
individual. Nonetheless, team spirit can play some role in fostering a kind of
collaboration through atmosphere and esprit de corps. In the Ryder Cup (a golf event
3. that involves American and European “teams”) the “collaboration is mostly of this ilk
although in some formats people may get some advice from their teammate and must take
into account, at least in a minor way, the particular skills of their partners. Notice
however, that the pace of golf allows such advice to be “out of time” from actual play and
to be quite explicit (e.g., “No, I wouldn’t try to hit a low iron out of that deep wet
rough.”) Games like hockey, soccer, American football and baseball require a high
degree of collaboration. However, the nature of the collaborative skills required is quite
different. The role differentiation is much higher in baseball and American football and
these roles often vary according to innate and learned skills as well. Even an extremely
good outfielder, for instance, might make a terrible pitcher. An excellent place-kicker
might be too small to be anything like an effective center. In role-playing games like
D&D and later, in on-line versions like WoW, the skill differentiation attempts to mirror
or even exceed the skill diversity found in the human populations. In typical professional
teams, the managers and coaches handle most of the skill and role differentiation through
explicit assignments to various people. However, even so, there are refinements to the
basic structure necessarily made by the team members themselves. For example, a third
baseman may know that the two first basemen on his team have different reaches and
throw accordingly. A catcher may see that a particular pitcher is not having much
success on a particular night with his fast ball and call for more breaking balls.
As these few examples show, the possibilities quickly seem to become too endless and
unwieldy to consider without an organized framework. In the next section we present
three simple frameworks that may help organize thinking about variations of
collaborative situations and skills. The first framework posits a simple metaphor for
thinking about and potentially measuring collaboration and the communication needed
for collaboration. The second presents and illustrates some factors of the environment
that tend to impact collaboration. (These are important as background to differentiate
performance due to situation from performance due to skill. The third framework posits
a limited set of important collaboration skills. These frameworks are clearly preliminary
and meant to stimulate dialogue; they are not meant as definitive nor based on a program
of formal empirical measurement.
A collaboration metaphor. Imagine two collaborators as two people moving through a
landscape more or less together. We call collaboration “good” to the extent that the
distance between them is minimized. This allows both instant and summative measures.
Their “productivity” is measured by their forward progress. (Depending on the situation,
progress may be joint, best, or least. We can also easily extend this metaphor to the N
person situation, but for purposes of exposition, consider the two person case). The two
collaborators are moving on opposite sides of a wall and potentially have different
landscapes to deal with as well as different skill levels. We can say that there are three
general classes of communication acts which will help maximize their progress and
collaboration. First, there are Stators in which one person describes the state of the
landscape as they can perceive it or their own internal state. Second, there are Turnors
which offer, describe, or command a change of direction. Third, there are Rators which
offer, describe, or command a change of rate. In various kinds of collaborative situations,
various means of communications are available. In ordinary conversation, these
4. collaborative signals are generally “metacomments.” For instance, a Stator about the
environment might be: “I am about to cover a very complex and confusing subject.” An
example of a Stator about internal state might be, “Whew. I am getting confused.” An
example of a very explict Turnor would be: “I think we’ve covered that topic enough.
Now, we need to consider….” One example of a Rator is simply backchannel nodding
and “right, right” said quickly which signals a desire to have the other person speed up.
A more explicit Rator comment might be: “Whoa! Hold on! What do you mean?” In
these examples, the communication is verbal, but of course, in various collaborative
situations, explicit hand-signals might be used or the collaborators might be expected to
rely directly on their own sensory input. A soccer player sprints forward quickly with the
ball and another player may take this as a signal to do the same although the first player’s
primary intention is to move forward past a defender not to do any signaling. The
“signaling” here is simply a by-product of the situation.
In general, collaborative situations will all require these types of communication events
in order to be maximally successful. However, the way in which these events are
communicated varies wildly from situation to situation. In many games, part of the rules
strictly limit how these events can be communicated. For instance, in bridge, it would
obviously be advantageous for the partners to show their hands to each other (or describe
them in detail). However, communication is limited to bidding (which has additional
constraints) and signaling out of this band via facial expressions, comments, etc. is
strictly forbidden. Similarly, charades and password are two collaborative games in
which the entire focus is really on communicating effectively in the face of strong
constraints on how that may be done.
Situational factors that impact collaboration. It is suggested that a five point scale is
sufficient to describe various situational factors. For example, we may ask about whether
coordinated rhythm is Required, Helpful, Neutral, Harmful or Incompatible with respect
to meeting goals. To win a gold medal in team rowing or ice-dancing, for instance,
coordinate rhythm is required. Achieving fine-scale rhythmic coordination, by the way,
is still a challenge for on-line environments.
We can ask the same set of questions (in the form of a five point scale) about other
aspects of a situation. For example, we suggest that asking these about conversation,
negotiation, shared gross stimulus context, shared fine stimulus context, and physical
contact. There are other aspects of the situation which do affect collaboration but cannot
be put so easily into this five point framework. For instance, the physical positioning of
people is important. Typically, having people “face each other” connotes opposition
while being shoulder to shoulder or in a circle connotes team membership.
In real situations, and especially in games, there are regulated and vital aspects of token
interaction. For instance, in soccer, American football, hockey, and basketball, there is
only one ball and possession of that ball is critical. This is reinforced by the fact that one
generally attempts to pass possession among one’s team members without letting any of
the members of the other team come into possession. In croquet, each person has their
5. own ball but these may interact. In fact, the strategy of croquet is mostly about these
interactions. By contrast, in golf, each person only hits their own ball.
A hierarchy of collaboration skills. Debates about the structure of human intelligence
remain despite a century of research and practice. At one extreme, some believe in a
“general” intelligence. Often this is differentiated into “fluid” and “crystallized”
intelligence, but others have suggested various numbers of components. At another
extreme, a strict behaviorist such as B.F. Skinner would basically focus on very specific
trainable skills and not focus on positing underlying abilities at all. This is not the place
to recount these debates in detail, but only to argue that such debates are often not to be
“settled” by empirical research but rather often reflect the aesthetic predispositions of the
investigators and/or their specific purposes.
In a similar vein, we may hypothesize that there may well be a general collaboration
factor. Some people seem to be better able (and/or desirous) to collaborate across a wide
variety of situations. On the other hand, whether or not one can effectively collaborate in
a given domain depends on some modicum of skill in that specific domain. An
individual may be a really good collaborator “in general” but if they are not athletic and
have never played soccer, then they will not be very effective as a collaborative soccer
player. They may not know what to “look for” how to translate from perceived situations
to appropriate actions or be able to carry out those actions.
This is an obvious but important point. There is still a danger in many actual on-line
collaborative games or virtual worlds that so much time and energy is required for a user
just to be minimally competent that learning about collaboration is quite secondary and
judging who is a “good collaborator” may depend far more on the game skill than on
whether they are actually a good collaborator.
Within the general skill of collaboration, we can conceptually separate skills for planning,
acting, and learning/adapting. Collaborative situations differ in terms of the degree of
preplanning that is possible, customary or necessary. Sometimes, in unplanned
emergency situations, a group of strangers is thrown together and must collaborate “on
the spot.” However, in most collaborative situations as diverse as professional sports,
and product development, a lot of time can and typically does go into planning the work
including how people will collaborate. In terms of acting in the moment, collaborative
skills could include noticing what others are doing, for deciding how best to react to the
on-going situation, and then for actually acting as well as for determining whether and
how pre-existing plans need to be changed. We could take any one of these skills and
break it down even further. For instance, the skill of noticing what others are doing
might be further analyzed according to the sense or senses involved. A separate set of
skills might be involved in groking patterns of actions. It seems clear even from such an
admittedly preliminary analysis that there are a wide variety of potential skills involved
in collaboration and no necessary a priori reason to expect a high degree of inter-
correlation based on ability or training. For example, someone who has learned to see
and react to patterns of play quite well in laser tag might be very inefficient coordinating
in WoW and vice versa.
6. Do on-line game collaborations transfer to real-life teamwork? This seems to be an
open question, but several of my colleagues; viz., Jason Ellis, Katherine Bessiere and
Wendy Kellogg are currently investigating whether real life distributed teams (of four)
can benefit from a series of three on-line games in Second Life. These three games
require increasing levels of coordination and collaboration. In one game, participant
teams work together moving Tetris-like pieces to form a bridge. In a second game,
participants attempt to build a tower a la the game “Blockhead.” In a third game,
participants are separated into sub-teams of “designers” and “implementers.” Designers
design a castle and implementers must build the castle so specified. However, there may
not be enough pieces to implement the original design.
Participants are given pre and post questionnaires to measure group cohesion and trust.
The teams engaging in the games are compared with a group who is simply given a
“space to hang out” in Second Life. In addition, the verbal and non-verbal behavior of
the participants during the collaborative games is being observed to determine, in an
exploratory way, whether it seems that there are particular types of experiences that seem
especially conducive to increased teamwork. Already it seems clear from observing
teams that whether or not they can “collaborate” effectively depends, among other things,
on whether they are proficient enough, after minimal training, to move, manipulate, and
communicate in Second Life.
References:
Jens Riegelsberger, M. Angela Sasse & John D. McCarthy (in press). Trust in Mediated
Interactions. In Katelyn McKenna, Tom Postmes, Ulf Reips , Adam N. Joinson (Eds.)
Oxford Handbook of Internet Psychology. Oxford: Oxford University Press.
Jens Riegelsberger & Asimina Vasalou (2007). Trust 2.1 - Advancing the Trust
Debate. Extended Abstracts of CHI 2007, San Jose, CA, US, April 28 - May 3.
U. S. Marine Corps (1994), Warfighting: The U.S. Marine Corps Book of Strategy. New
York: Currency Doubleday.