CAR-TR-826 May 1996
CS-TR-3637
Human-Computer Interaction Laboratory
Center for Automation Research
University of Maryland, College Park, MD
20742-3255
Abstract
To take advantage of todayís technology, many organizations are migrating from their legacy systems. With help from the Human-Computer Interaction Laboratory (HCIL) and Cognetics Corporation, the Maryland Department of Juvenile Justice (DJJ) is currently undergoing an effort to redesign their information system to take advantage of graphical user interfaces. As a research lab, HCIL identifies interesting research problems and then prototypes solutions. As a project matures, the exploratory prototypes are adapted to suit the end product requirements. This case study describes the life cycle of three DJJ prototypes: (1) LifeLines, which uses time lines to display an overview of a youth in one screen, (2) the DJJ Navigator, which helps manage individual workloads by displaying different user views, and (3) the ProgramFinder, a tool for selecting the best program for a youth.
INTRODUCTION
Social service agencies are often plagued with legacy software
systems that are difficult to use and provide inadequate functionality.
One cause of these problems is poorly designed user interfaces.
A good user interface can increase performance, reduce errors,
and enhance user satisfaction. Advances in user interface technology
coupled with the affordability of personal computers are motivating
many organizations to redesign their outdated systems (Sawyer,
1995).
In 1994, the Human-Computer Interaction Laboratory (HCIL) began working with the Maryland Department of Juvenile Justice (DJJ) in an effort to redesign the user interface of their current information system, ISYS (Information System for Youth Services). ISYS is a terminal based system used to handle 50,000 juvenile complaints per year (Figure 1). It is used by about 600 DJJ employees in various offices and facilities across the state.
NISYS, the next generation ISYS, will run on PCs in a WindowsTM
environment. Visualization techniques are used to display youth
record overviews, and worker-specific screens facilitate task
scheduling and document management.
The Redesign Effort
To learn and assess ISYS, we conducted 22 field visits, attended training sessions, and administered the Questionnaire for User Interaction Satisfaction (QUIS) (Rose, Shneiderman & Plaisant, 1995; Slaughter, Norman & Shneiderman, 1995; Vanniamparampil, Shneiderman, Plaisant & Rose, 1995). Based on our findings, we proposed 28 short term recommendations to improve ISYS while NISYS was being developed. Our recommendations focused on making system access easier, improving data accuracy, making information retrieval easier, increasing the usefulness of the system, and enhancing user satisfaction. DJJís initial response was to take action on all 28 recommendations. However, because of the effort dedicated to designing NISYS only a subset of the recommendations have been implemented to date.
Next we developed three NISYS prototypes:
ï LifeLines, which uses time lines to display a youthís history with DJJ in one screen (Plaisant, Milash, Rose & Widoff, 1995; Milash, Plaisant & Rose, 1995),
ï the DJJ Navigator, which helps manage individual workloads by displaying different user views, and
ï the ProgramFinder, which can be used to find the best program
for a youth.
Early prototypes have been shown to approximately 60 DJJ personnel,
ranging from case workers to supervisors. Overall their reactions
have been very positive. We are refining the prototypes and integrating
them into NISYS based on their feedback.
In 1995, Cognetics Corporation joined our team. HCIL and Cognetics
are designing NISYS jointly using Cogneticsí Design Methodology
(CDM) for quality usability engineering (Kreitzberg, 1995). Cognetics
is responsible for preparing the request for proposal (RFP) for
NISYS and for developing a testable prototype. The RFP will contain
the user interface specifications in addition to the functional
specifications.
In addition to redesigning the user interface, it has become evident
that many of DJJís processes need to be re-engineered.
Specifically, the role of documents in process re-engineering
has emerged as a major focus of this project (Greene, 1996).
We compiled a comprehensive inventory of DJJ documents, analyzed
their reporting needs, and are currently helping them redesign
their documents. Many of the new documents will be incorporated
into NISYS.
LIFE CYCLE OF USER INTERFACE TECHNIQUES
In this paper, we describe the life cycle of three NISYS prototypes
from problem to product (Figure 2). The cycle begins with a problem
that needs to be solved. Exploratory prototypes, ranging from
paper mockups to Delphi applications, are then developed to illustrate
possible solutions and communicate ideas among the designers,
users, and implementors. For researchers, the next stage typically
involves generalization or conceptualization of the techniques
used. This is when most of the research papers are written.
The other path is more commercial. It leads to the refinement
of the prototypes and integration into a final product. HCILís
primary role in the NISYS project has been to design exploratory
prototypes for key functional areas while Cognetics is responsible
for integrating these ideas into the NISYS design.
The exact sequence of these stages varies from project to project
(Figures 3, 6, and 11). The white ovals represent phases accomplished
during the NISYS project, the gray ones were by done previously,
and the hashed ones remain to be done. The LifeLines and Navigator
designs started at the problem stage while the ProgramFinder is
an application of a general visualization technique that had been
developed in a previous project. The NISYS project is ongoing
so the product designs are not final.
LifeLines
Problem:
With ISYS, it is very difficult and time consuming to get an overview
of a youthís history with DJJ. Case workers have to use
cryptic codes to navigate through dozens of screens. Important
information is difficult to find, data entry problems are common,
and processing delays often go undetected. During an ISYS training
session we were told that ìthe trick is to find the magic
case,î i.e., you have to be lucky enough to find the case
that contains the information you need.
Exploratory Prototypes:
LifeLines uses multiple timelines to present a youth record overview
in one screen. The buttons on the top provide access to general
information, such as aliases and significant others. Critical
flags (e.g., suicide risk) are displayed above the buttons. All
of the cases, placements, assigned workers, and reviews are plotted
on timelines. Each facet can be expanded or collapsed (i.e.,
hide the labels and draw the lines closer together).
The timelines provide a visual overview of the youth record.
From Bartís record (Figure 4), we can see that he has been
involved with DJJ for slightly over a year. During that period,
he has been referred four times, twice for breaking and entering
(B&E), once for auto theft, and most recently for attempted
murder. Line thickness is used to indicate severity and color
indicates the depth of penetration into the system. The two breaking
and entering cases were handled informally by DJJ (no court intervention)
and the auto theft case was sent to court where Bart was found
delinquent (DELQ). As a result, Bart was committed to Cheltenham
(a residential youth facility). The right edge of the timelines
shows Bartís current status: he is detained at Waxter (a
detention center) while case manager Brown decides whether to
forward the attempted murder charge to court.
Detailed information can be retrieved with a click or two of the
mouse. A click on a review icon brings up the text of the review.
Similarly details about the cases or placements can be obtained
by clicking on the lines or labels. Clicking on the workerís
name brings up contact information. Relationships among the different
timelines can also be highlighted. For example, one cannot tell
by looking at the timelines if the drug abuse program was a result
of the auto theft case or the breaking and entering case, but
clicking on the drug abuse program highlights the related case
(the B&E case in this instance) and the worker assigned to
that case.
DJJís reaction to LifeLines has been very positive. Users
especially like how easy it is to access details and get a quick
overview of a youth. The timelines also help users discover data
entry problems, like cases that were never closed. This type
of error normally often goes undetected in ISYS. A few users
have expressed concerns about the possible bias associated with
color and thickness coding (Shneiderman & Rose, 1996).
Generalization:
Following the exploratory prototype, the HCIL team explored other
application domains and extended the LifeLines to a general technique
for visualizing personal histories, such as medical, financial,
legal, and education histories. For example, the facets of a
patient record might include conditions, medications, and doctorís
visits. The lines within each facet represent stories, like a
three year history of heart problems. Each story may also have
a series of episodes or periods, such as light versus acute back
pain, which are shown using line thickness and color. Discrete
events, like doctor visits or legal decisions, are shown with
icons. LifeLines has been applied to a sample cardiology patient
record (Plaisant & Rose, 1996).
Product:
Based on suggestions from DJJ and on the learning from the generalization
phase, LifeLines is being refined and incorporated into the NISYS
Youth Record (Figure 5). We believe that the successful completion
of the generalization phase, followed by specialization of the
LifeLines, was a very beneficial process which lead to a more
complete and flexible interface.
The major difference is the addition of the event list. While
LifeLines is very useful for showing the current status of a youth,
highlighting relationships, and spotting anomalies, the event
list is better for showing specific dates and more textual detail.
The two displays are tightly coupled. Clicking on an event in
the timelines highlights it in the event list, and vice versa.
Double clicking on an event in either display brings more details.
Users can also filter the list by type (e.g., reviews, placements).
To help users focus their attention on important information,
green is used to indicate normal events and red signals critical
events. To facilitate workflow, future events are displayed and
popup menus accessible from the line segments provide access to
the documents associated with different processes. A click on
a future review pops up a blank review form and a click on a caseís
intake period (light blue section) displays a menu of intake documents.
The primary uses of the Youth Record will be to get overviews
of new youths and to work on current assignments. The Today button
allows users to quickly zoom in on the active cases.
Different types of users are interested in different information.
For example, a nurse wants to see medical information while a
teacher is more interested in the education history. Therefore,
a popup control panel was added to allow users to customize the
contents of the timelines.
Specific rules for label abbreviation and positioning have been
designed. Other aspects specific to the application are still
being worked on (e.g., how to display cases that have been merged
or split by the courts).
DJJ Navigator
Problem:
One of the key problems with ISYS is that users must search through
thousands of records simply to find the information they are interested
in. Supervisors, case managers and facility workers are all interested
in different information. The information is there but is not
presented in a manner which facilitates the usersí work.
Many workers see the system as a burden rather than an aid.
They perceive ISYS as a sort of black hole that they must put
data into while receiving little back for their effort.
Exploratory Prototypes:
The DJJ Navigator helps manage individual workloads by displaying different user views. For example, a facility worker needs to view the youths in their facility while a case manager would be interested in seeing only their caseload. In general, DJJ works with the youth as a whole or with an individual case involving a youth. Because of this distinction, the first Navigator prototype (Figures 7 and 8) used a multi-paned approach to show youth and case data. Additional data (e.g., victims) could be shown in another pane.
The Headquartersí View provides an overview of all the
cases and youths currently being handled by DJJ. The cases are
divided by area of jurisdiction and the youths by type of supervision.
Selecting different levels of information displays different
views. For example, doubling clicking on PG in Area V would display
the Prince Georgeís Office View. The cases would be divided
by units within the office (e.g., intake, investigation, case
management) and only youths being supervised by PG workers would
be displayed. Users can also use controls to focus on specific
data (e.g., cases with pending intake decisions or youths who
are currently AWOL).
A second prototype was developed that focused more specifically
on DJJís workflow by providing a group of pre-defined user
views to support different tasks (Figure 9). The emphasis moved
from user manipulation of views to user selection of preset views
and the multi-paned approach was replaced by a single spreadsheet-like
display. A case worker would have a To Do view which would display
a list of things they needed to do (e.g., schedule intake hearings,
make intake decisions, etc.). Clicking on an item in the list
would popup the paperwork partially filled out. More experienced
users would be able to create their own custom views to help manage
their workloads. They could display different type of elements,
like cases, youths, victims, or events.
Generalization:
We have not explicitly generalized the Navigator to other applications
but we can imagine using it as a generic workflow or personal
role management tool (Shneiderman & Plaisant, 1994; Landsdale,
1988). The first Navigator could be used to display and navigate
hierarchical, relational datasets. For a school, cases and youths
could become students, courses, and teachers.
To support generic datasets in the first Navigator (Figures 7
and 8), we proposed the Pane Definition Language (PDL), a formal
language which allowed users to specify Navigator views and relationships.
This language proved inadequate when there are dozens of views
because describing each one independently is time consuming and
leads to potential inconsistencies. Our second attempt was a
rule-based scheme which endeavored to define a concise set of
rules for generating the views. This also proved problematic
as some views needed to be tailored to specific tasks and did
not fit our rules. Finally, we started investigating a hybrid
approach that would consist of a general rule set as well as some
PDL-like screen descriptions. This approach would allow the user
to address special cases while doing away with the need to describe
every screen.
Product:
Without a good general model of the Navigator user interface we
had to move toward very specific individual Navigator screens
custom tailored for each type of user. These screens will be
the first screen a user sees when they log on to NISYS. Current
efforts are focused on the Case Worker screen and defining the
tools needed in this domain (Figure 10). The workerís
case assignments plus a set of preset views are still shown in
a spreadsheet display. The contact and to do lists have been
given their own areas since his information is used frequently.
We have tried to propagate frequently used youth information
to the Case Worker screen so the time spent swapping between screens
is minimized. Hooks to other applications, like word processors
and email, were also added.
We plan to continue refining the Case Worker View and delve further
into defining other user views such as clerical and facility workers.
Here the lack of generalization principles limits our ability
to rapidly prepare screens for all the different users. On the
other hand, it is likely that the continuing design of these screens
will sharpen our understanding of the interface challenges and
lead us toward general principles for the design of case management
user interfaces.
ProgramFinder
Problem:
DJJ personnel, newspaper reporters and other social service agencies
often need answers to questions like ìHow many auto thefts
were committed last month?î, ìWhat is the best program
for this youth?î, or ìWhich offenses are typically
committed by 11 to 13 year olds?î Answering these questions
may involve writing a computer program to extract data from ISYS
or manually searching through hundreds of screens or paper files.
Both of these techniques are very time consuming and often lead
to incorrect answers. NISYS needs to allow users to explore the
data and make queries.
Exploratory Prototypes:
Database query languages with tabular output or spreadsheets could
satisfy many of DJJís standard needs but we suggested the
use of data visualization tools as an appealing and powerful alternative.
By visualizing data and allowing fast and easy exploration, dynamic
query (DQ) applications help users identify outliers, explore
trends, and support routine and ad-hoc queries. Dynamic queries
continuously update search results in a visual display (e.g.,
scatterplot, map, etc.) as users make queries by adjusting sliders
or selecting buttons (Williamson & Shneiderman, 1992; Ahlberg
& Shneiderman, 1994).
We used IVEE to demonstrate DQ applications to DJJ. The Information
Visualization and Exploration Environment (IVEE) is a product
based on earlier dynamic query research. IVEE is a generic tool
that automatically builds DQ applications for a given dataset
by creating query controls (alpha sliders, range sliders, or toggle
boxes) for each attribute and visualizing the elements in a starfield
display (xy scatterplot) or on a map (Ahlberg & Wistrand,
1995).
First, a sample dataset of 4800 case referrals was visualized
and explored (Figure 12). Two interesting trends were discovered
immediately: the ages for several youths were incorrect (e.g.,
95 years old, 35 years old, less than one year old) and many intake
decisions required more than the 25 day maximum. Upon further
exploration, we discovered several patterns of referrals, including
one that indicated a drug bust (Rose & Vanniamparampil, 1995).
DJJ found it very appealing to be able to explore trends in their
data and make queries so easily so we started exploring other
potential applications. Using IVEE, we demonstrated the Dynamic
ProgramFinder (Figure 13), a tool for selecting the best programs
for a youth based on criteria such as security, location, and
services provided.
Generalization:
In this case, the generalization phase took place before the DJJ
project even started. We could follow visual information seeking
principles (Ahlberg & Shneiderman, 1994) and use a product
outcome of the research to generate the next round of exploratory
prototype.
Product:
DJJ can use IVEE to support their research and general analysis
needs but some parts of NISYS were found critical enough to require
the design of a customized DQ application (e.g., a ProgramFinder
to find the best programs for the youths). By including the ProgramFinder
in NISYS, it may also be possible to automatically set some of
the query criteria based on prior youth assessments. The NISYS
version of the ProgramFinder (Figure 14) provides access to individual
Youth Records and allows users to select a set of programs and
then send a referral packet to each. A referral packet is the
set of documents sent to a program requesting that they consider
admitting a youth.
The custom ProgramFinder can display the programs on a map or
in a spreadsheet. The spreadsheet has the advantage of being
able to show more program attributes at once while the map is
better for spatial information. In the map display, shape indicates
the type of program (e.g., residential or non-residential) and
color shows how well a program matches the settings. The house
icon shows where the youth lives so transportation and visitation
issues can be considered. The program details were included in
the side window.
To avoid scrolling lists of controls, tab panels were used to
divide the controls into smaller more manageable groups: general,
restrictiveness, and intervention. The restrictiveness controls
presented a challenge. In addition to specifying an acceptable
range of values, DJJ also wanted to be able to specify the ideal
value. The ideal value along with an assigned weighting determines
how well a program matches. The range sliders were modified to
allow users to specify an ideal value by clicking on it. Yellow
stars mark the ideal values and black vertical lines mark the
values of the selected program.
The criteria that should be used when choosing an appropriate
program are still an area of debate. DJJ is also interested in
extending the ProgramFinder to monitor program acceptances and
rejections. IVEE could be used to analyze the data and determine
the kinds of programs that are needed.
CONCLUSION
The NISYS project is an example of a successful collaboration
among two state agenciesóthe University of Maryland working
with the Maryland Department of Juvenile Justice. The University
has been able to offer DJJ objective solutions to their user interface
problems, unbiased by commercial agendas. By demonstrating our
exploratory prototypes early on, we were able to help motivate
DJJ to design a new system. We also acted as instructors and
facilitators - teaching DJJ the basics about graphical user interfaces,
networking, and database design, in order to put them in a better
position to manage the implementors of NISYS.
The academic-industry partnership between HCIL and Cognetics is
the other contributing factor to this projectís success.
There are two main paths in the life cycle of a user interface
technique: one that is more research oriented and one that is
more commercial. While Cognetics follows the product path and
works on the overall product design, HCIL can focus on the research
issues and explore key functional areas, like LifeLines. Working
on real problems generates ideas and participation in the commercialization
of these ideas helps refine the principles and produce more general
designs. Three of HCILís research prototypes have been
refined and integrated into the NISYS design.
By including the user interface design in the RFP, DJJ hopes to
better convey their needs to potential bidders plus the design
can be used to hold implementors accountable for what they deliver.
Designing the system up front will also help avoid implementation
biases and ensure that the resulting system suits DJJís
needs. Many detailed decisions will still be left to the implementors
so basic guidelines for making additions and modifications to
the design will be included in the RFP.
ACKNOWLEDGMENTS
We thank Chris Cassatt (DJJ) and the Cognetics team, Charles Kreitzberg,
Scott Gilkeson, and Jim Kauffman, for their contributions to the
NISYS project. We also thank Ben Shneiderman for his early contributions
to the project. The preparation of this report was supported
by funding from the Maryland Department of Juvenile Justice.
REFERENCES
Ahlberg, C., Shneiderman, B., (1994), ìVisual Information
Seeking: Tight Coupling of Dynamic Query Filters with Starfield
Displays,î Proceedings of Human Factors in Computer Systems
ë94 (Boston, Mass., April 1994), ACM, New York, 313-317.
Ahlberg, C., Wistrand, E., (1995), ìIVEE: An Information
and Visualization & Exploration Environment,î Proceedings
of IEEE Visualization ë95 (Atlanta, Georgia, October 1995),
IEEE Computer Society, Los Alamitos, CA, 66-73.
Greene, S., (in preparation), ìInformation and Process
Integration From User Requirements Elicitation: A Case Study of
a Social Services Agency,î HCIL Technical Report, University
of Maryland, College Park, MD.
Kreitzberg, C., (1995), ìManaging for Usability: The Cognetics
Design Methodology,î Cognetics, Corp., Princeton Junction,
NJ.
Landsdale, M., (1988), ìThe psychology of personal information
management,î Applied Ergonomics, 19(1), Butterworth
& Co. Ltd., 55-66.
Milash, B., Plaisant, C., Rose, A., (1995), ìLifeLines:
Visualizing Personal Histories,î HCIL 1995 Video Reports,
University of Maryland, College Park, MD.
Plaisant, C., Milash, B., Rose, A., Widoff, S., Shneiderman, B.,
(1995), ìLifeLines: Visualizing Personal Histories,î
Proceedings of CHI ë96 (Vancouver, BC, April 13-18 1996),
ACM, NY, 221-227.
Plaisant, C., Rose, A., ìExploring LifeLines to Visualize
Patient Records,î HCIL Technical Report, CAR-TR-819,
University of Maryland, College Park, MD.
Rose, A., Shneiderman, B., Plaisant, C., (1995), ìAn Applied
Ethnographic Method for Redesigning User Interfaces,î Proceedings
of DIS ë95 (Ann Arbor, Michigan, August 1995), ACM, NY,
115-122.
Rose, A., Vanniamparampil, A., (1995), ìUsing Dynamic Queries
for Youth Services Information,î HCIL 1995 Video Reports,
University of Maryland, College Park, MD.
Sawyer, P., Mariani, J., (1995), ìDatabase systems: challenges
and opportunities for graphical HCI,î Interacting with
Computers, 7(3), Butterworth-Heinemann, 273-303.
Shneiderman, B., Plaisant, C. (1994), ìThe Future of Graphic
User Interfaces: Personal Role Managers,î People and
Computers IX, British Computer Societyís HCI ë94
(Glasgow, Scotland, August 1995), 3-8.
Shneiderman, B., Rose, A. (1996), ìSocial Impact Statements:
Engaging Public Participation in Information Technology Design,î
Proceedings of Computers and the Quality of Life ë96 (Philadelphia,
PA, February 14-15, 1996), ACM, NY, 90-96.
Slaughter, L., Norman, K., Shneiderman, B., (1995), ìAssessing
Usersí Subjective Satisfaction with the Information System
for Youth Services (ISYS),î Proceedings of Third Annual
Mid-Atlantic Human Facotrs Conference (Blacksburg, VA, March 26-28,
1995), 164-170.
Williamson, C., Shneiderman, B., (1992), ìThe Dynamic HomeFinder:
Evaluating Dynamic Queries in a Real-Estate Information Exploration
System,î Proceedings of ACM SIGIR ë92 (Copenhagen,
June 21-24, 1992), ACM, New York, 338-346.
Vanniamparampil, A., Shneiderman, B., Plaisant, C., Rose, A., (1995), ìUser Interface Reengineering: A Diagnostic Approach,î HCIL Technical Report, CAR-TR-3459, University of Maryland, College Park, MD.