Investigations into Team Decision Making Processes
TITAN was developed as a vehicle for investigating decision making processes in
hierarchically organized teams. It is used to assist in developing models relating
A typical experiment involves a group leader and up to three subordinates.
Subordinates can be real participants or can be simulated. All players are presented
with a radar view and targets.
- How individuals collect and assess information
- How leaders weigh the inputs of subordinates
- How leaders integrate assessments provided by subordinates
- How feedback affects these issues
- How confidence is affected by short and long term performance
Each subordinate can typically access a subset of target characteristics. S/he uses
this to assign a threat level that, once entered, is reflected on the leader's
display. The leader is responsible for assessing an over all threat level by
evaluating assessments submitted by subordinate(s). Once this has been done, the
leader may be presented with feedback (error margin of the last assessment, error
percentage over number of assessment, error made by each participant). This feedback
enables the leader to adjust his or her interpretation of subordinates work.
The 'game' has two main configurations. It can be run as a stand-alone or network
Single mode consists of one player. In this setup the leader may be responsible for
assessing all target information. Also, one or more simulated agents may be added.
Simulated subordinates would behave as real people as far as the leader is concerned.
Their behavior, however, is tightly controlled (eg. their response time, and error
Team mode consists of a leader and up to 3 subordinates, some of whom may be agents.
It is run typically over LAN but, as it is TCP/IP compliant, experiments were also
conducted over the Internet.
Different experimental protocols are specified through scripts. These allow the
experimenter to vary...
Thus stand alone and team experiments can be created in a number of significantly
different variations without any programming changes.
- The menu structure presented to each player
- The assignment of target characteristics to subjects
- Size and placement of GUI components
- Level of feedback provided to each player
- Presence or absence of historical performance information
- Whether or not confidence testing is performed on each player
- The number of real and simulated players
- Error and response time characteristics for simulated players
- Level of detail recorded by the data logger
Object Oriented Design
TITAN's design relies extensively on the MVC (Model-View-Controller) paradigm of
object oriented programming. All components are independent, detached, flexible.
Object Oriented architecture enables efficient reusability and adaptation of the
system to frequently changing experimental requirements
most of the systems' modules (objects) are loaded and configured at run-time
(Dynamic Binding). This approach, again, provides the system with much enhanced
Threat assessment values (may come from a real participant, a simulated player, a
network proxy) are passed to a threat model. Views are then notified by models
(typically by an observation mechanism). Views are responsible only for visual
presentation of threat assessments. Their number, size and location on the screen,
finer details etc. are provided by scripts when the experiment is born. Models are
totally insensitive to the number and nature of views that observe them.
TITAN consists of over 150 classes (approx. 20,000 lines of code).
all programming was done in JAVA and the distributed part of the application
(network connection between leader and subordinates) was developed with RMI (Java
Remote Method Invocation). It makes TITAN a truly platform independent application
that can be run over the Internet from virtually any workstation.
Other JAVA projects
Other Human Performance projects
Please visit our home page:
NTT Systems Inc.