SUPPLE: Automatic Generation of Personalizable User Interfaces
Last updated: August 12, 2009
August 28, 2009:
This page is no longer actively maintained. To view a short video demonstrating Supple's capabilities, to read about recent developments, or to see a list of media mentions, the the new site.
Today's user interfaces are typically designed with the assumption that they are going
to be used by an able-bodied user, who has typical perceptual and cognitive abilities, who
is sitting in a stable warm environment, and who is using a typical set of input and output
devices. Any deviation from these assumptions (for example, hand tremor due to aging, low
vision, riding on a jostling bus, trying to use a laser pointer to control a mouse cursor, or trying to access an application from a mobile device), may
drastically hamper users' eectiveness--not because of any inherent barrier to interaction,
but because of a mismatch between users' situation and the assumptions underlying
the user interface design.
We argue that interfaces should be personalized to better
suit the contexts of individual users. Many personalized interfaces are needed because of
the myriad of distinct individuals, each with his or her own abilities, preferences, devices
and needs. Therefore, traditional manual interface design and engineering will not scale to
such a broad range of potential contexts and people. A dierent approach is needed. We believe that automatically generated user interfaces, which are adapted to a person's
devices, tasks, preferences, and abilities, can improve people's satisfaction and
performance compared to traditional manually designed "one size ts all"
As part of the SUPPLE project, we have developed three systems to enable a broad range of personalized
adaptive interfaces: SUPPLE itself, which uses decision-theoretic optimization to automatically
generate user interfaces adapted to a person's device and usage; ARNAULD, which allows
optimization-based systems to be adapted to users' preferences; and the ABILITY MODELER,
which performs a one-time assessment of a person's motor abilities and then automatically
builds a model of those abilities, which SUPPLE uses to automatically generate user interfaces
adapted to that user's abilities. The results of our laboratory experiments show that
these automatically generated, ability-based user interfaces signicantly improve
speed, accuracy and satisfaction of users with motor impairments compared to
manufacturers' defaults. We also provide the first characterization of the design space of
adaptive graphical user interfaces, and demonstrate how such interfaces can signicantly
improve the quality and e
ciency of daily interactions for typical users.
In addition to our other publications listed below, our AAAI'08 paper provides a 5 page overview of the main contributions of the project and Krzysztof's dissertation provides the details.
A version of the SUPPLE is Aavailable for download.
The publications are listed in chronological order (most recent at the bottom) and the key papers are highlighted.
- The underlying vision of adaptive user interfaces; the paper
describes a number of earlier projects that lead to SUPPLE as well
as the basic assumptions of the SUPPLE system:
- D. Weld, C. Anderson, P. Domingos, O. Etzioni, T. Lau, K. Gajos, and S. Wolfman.
Personalizing User Interfaces In Proceedings of
The "original" SUPPLE paper describing the models, the casting
of UI rendering into an optimization problem and the
- Krzysztof Gajos and Daniel S. Weld.
SUPPLE: Automatically Generating User Interfaces. In Proceedings of
IUI'04. Funchal, Portugal, 2004
An extended abstract from a workshop
describing practical issues involved in using SUPPLE to generate user
interfaces for ubicomp applications. This paper has been subsumed by our Ubicomp'05 paper.
Krzysztof Gajos and Daniel S. Weld. Automatically Generating User Interfaces For Ubiquitous Applications. In Workshop on Ubiquitous Display Environments, Nottingham, UK, 2004.
An extended abstract outlining recent progress
on customization, automatic adaptation
and elicitation of the parameters for the cost functions that
guide the UI generation process (it has been subsumed by our Ubicomp'05 and UIST'05 papers):
Krzysztof Gajos, Raphael Hoffmann and Daniel S. Weld. Improving
User Interface Personalization. In UIST'04. Santa Fe, New Mexico,
Another workshop paper describing
our approach to maintaining consistency in how the user
interface for an application is rendered on different devices.
Krzysztof Gajos, Anthony Wu and Daniel S. Weld. Cross-Device Consistency in Automatically Generated User Interfaces. In Workshop on Multi-User and Ubiquitous User Interfaces (MU3I'05). San Diego, CA,
This Ubicomp paper elaborates a number of practical issues such as the extensions ot the interface modeling language, an adaptation algorithm with a supporting user study, the customization facitlity and the implementation details including support for distributed operation.
Krzysztof Gajos, David Christianson, Raphael Hoffmann, Tal Shaked, Kiera
Henning, Jing Jing Long, and Daniel S. Weld.
Fast And Robust Interface Generation for
In Proceedings of the Seventh International Conference on Ubiquitous Computing (UBICOMP'05). Tokyo, Japan, September, 2005.
This UIST paper presents a
new system, Arnauld that interactively
elicits user preferences for the purpose of automatically learning
parameters of optimization-based systems. We have used it to
automatically generate the parameters of SUPPLE's cost functions.
Krzysztof Gajos and Daniel S. Weld. Preference Elicitation
for Interface Optimization. In Proceedings of UIST'05, Seattle,
WA, USA, 2005.
This paper attempts to identify those aspects of adaptive interfaces
that make some of them a pleasure to work with while others are
annoying hinderances. We designed three different adaptive GUIs and
evaluated them in two different experiments. Finally, we synthesized
our results with those of past studies and identified someof the
design factors that appear to contribute the most to an adaptive
Krzysztof Z. Gajos, Mary Czerwinski, Desney S. Tan and Daniel S. Weld. Exploring the Design
Space For Adaptive Graphical User Interfaces. In Proceedings of
AVI'06, Venice, Italy, 2006.
This short abstract introduced the idea of using SUPPLE to automatically generate user interfaces adapted to people's abilities. The next paper realized this vision.
Krzysztof Z. Gajos, Jing Jing Long and Daniel S. Weld
Generating Custom User Interfaces for Users With Physical
Disabilities In Proceedings of ASSETS'06, Portland, OR, 2006.
In this paper we present a novel way to model the motor abilities of users with atypical devices and abilities. We then develop a new optimization-based algorithm that allows SUPPLE to automatically generate user interfaces adapted to people's individual motor and vision abilities.
Krzysztof Z. Gajos, Jacob O. Wobbrock and Daniel S. Weld. Automatically Generating User Interfaces Adapted To Users' Motor And Vision Capabilities. In Proceedings of UIST'07, Newport, RI, USA, 2007.
This workshop paper, written in collaboration with the Change Group explores the potential impact of using automatically generated and adaptive user interfaces in health care delivery in developing regions.
Brian DeRenzi, Krzysztof Gajos, Tapan Parikh, Neal Lesh, Marc Mitchell, and Gaetano Borriello. Opportunities for Intelligent Interfaces Aiding Healthcare in Low-Income Regions. In Proceedings of IUI4DR - Workshop on Intelligent User Interfaces for Developing Regions (at IUI'08). Canary Islands, Spain, 2008.
The results of this study show that people with motor impairments are faster, more accurate, and strongly prefer automatically generated ability-based user interfaces to the manufacturers' baselines.
Krzysztof Z. Gajos, Jacob O. Wobbrock and Daniel S. Weld. Improving the performance of motor-impaired users with automatically-generated, ability-based interfaces. In Proceedings of CHI'08, Florence, Italy, 2008.[Best Paper Award]
This paper contributes to the systematic exploration of the design space of adaptive user interfaces. Our study explores the relative importance of accuracy and predictability of the predictive algorithm driving the adaptation process.
Krzysztof Z. Gajos, Katherine Everitt, Mary Czerwinski, Desney S. Tan and Daniel S. Weld. Predictability and accuracy in adaptive user interfaces. In Proceedings of CHI'08. Florence, Italy, 2008. CHI Note.
This is a 5 page summary of all the main contributions of the SUPPLE project.
Krzysztof Z. Gajos, Daniel S. Weld, and Jacob O. Wobbrock.
Decision-Theoretic User Interface Generation.
In Proceedings of AAAI'08, NECTAR paper track. Chicago, IL, USA. 2008.
Krzysztof's dissertation describes project's contributions in detail.
Krzysztof Z. Gajos. Automatically Generating Personalized User Interfaces. Ph.D. Dissertation. 2008.