NO.193 The Moving Target of Visualization Software - Closing the Gap between Research and Application
February 12 - 16, 2024 (Check-in: February 11, 2024 )
- Michael Krone
- University of Tübingen, Germany
- Christina Gillmann
- University of Leipzig, Germany
- Takayuki Itoh
- Ochanomizu University, Japan
- Alexander Lex
- University of Utah, USA
- Guido Reina
- University of Stuttgart, Germany
Visualization has evolved into a mature subfield of computer science. It has become widely accepted as an important data analysis method in diverse fields. Visualization not only enables scientific data analysis, but is also widely adopted in industry, as the success of business data visualization platforms such as Microsoft PowerBI or Tableau demonstrate. Visualization is also widely used in the news media to communicate data about topics as diverse as the spread of COVID-19 or the uncertainty associated with political polls. Over the years, the academic visualization community has developed many tools and methods, many of which have been widely adopted in all of these areas. However, many research prototypes never reach the maturity necessary for broad adoption, even though the underlying method has significant merit. These prototypes are often neither sustainable nor easily extensible for subsequent research projects, and their lifespans are often tied to the original author’s academic career. Since research prototypes in academia are usually implemented by PhD students as part of their thesis projects, this results in a rather short lifespan. Another aspect that adds to this problem is that the whole ecosystem around a software quickly evolves. This for example includes changes in the execution environment (software as well as hardware), preferred programming languages, external third-party libraries, and interaction paradigms. Consequently, PhD students would have to invest a lot of time and effort to keep up with the moving target of developing and maintaining a usable software, while of course also doing actual research at the same time.
We found this topic to be of great interest to many members of our community, which is the reason why we have recently organized a number of related events to discuss the most pressing issues and propose medium-term solutions that can improve both the practical aspects of our daily work and the quality and efficiency of our research contributions. The first event in this series has been the Shonan Seminar #145 (February 2019, organized by H. Childs, T. Itoh, M. Krone, G. Reina), in which the group (consisting of 24 participants from industry, national labs, and academia) distilled the nine most interesting concerns and opportunities. We have meanwhile transformed these results into a publication that shows the state of our community as well as possible future directions.
In this seminar, we want to bring together leading visualization researchers and practitioners to discuss the specific challenges around visualization software. We identify the following challenges that we want to work on during this seminar. Visualization research usually requires an implementation of the proposed method or approach for evaluation. However, the visualization community lacks incentives to publish research software, much less evolving a research prototype into a usable tool. Furthermore, especially young researchers usually lack the training and experience to develop sustainable research software. As mentioned above, the requirements for novel approaches are steadily increasing due to data scale and complexity, but also due to the increasing maturity of the field itself and the influences from other areas, such as machine learning or human-computer interaction. Consequently, the prototypes usually are not easily extensible for subsequent research projects. In order to close the gap between research and application, we want to develop proposals and strategic initiatives to solve these challenges and to move the whole community forward.
The aim of the proposed Shonan Seminar is to intensify and advance the discussions started at Shonan Seminar #145 by elaborating on the yet unfinished discussions and to discuss additional topics, as suggested in the following. One new aspect beyond the previous seminar would be the discussion of concrete models for incentivizing visualization software within the research community (e.g. tools papers, which are common and considered important contributions in other scientific disciplines, mandatory software accompanying submissions, etc.). In addition, we aim to specify requirements and changes in the VIS community that need to be fulfilled in order to ease the translation of research prototypes into applicable software.
Topics to Be Discussed in the Seminar
We identified five challenges for software development that are specific to visualization research, which we want to work on during this seminar:
- The lack of incentives and opportunities offered by the visualization community to do systems-heavy research or at least develop a research prototype into a usable tool.
- The fundamental problem of increasing complexity and requirements of systems due to also increasing data scale and complexity, but also due to increasing expectations, and the influences from other areas, such as machine learning or HCI.
- The challenge of collaborating with domain scientists and having visualization tools adopted by them in their daily routine. This also includes the challenge of establishing best practices for the collaboration where both sides benefit.
- Lack of unified software in visualization research, which often results in starting from scratch, and the lack of platforms to collaborate.
- Lack of community-accepted rules for publishing software or code when publishing papers about new methods or applications in visualization.
- Formation of a group that elaborates a typology of features and design choices for visualization software, with the goal of a journal publication and launching a curated online database.
- Definition of a lightweight interface protocol for coupling visualization systems. Formation of a subgroup that will steer the specification and open source example implementations.
- Development of guidelines for the visualization community to reconcile research contribution and reproducibility of research via availability of respective software, similar to other fields in CS.
- Build an international network of experts for the diverse available visualization software that work on a curriculum, which would allow for a structured training of students and young researchers in the field of visualization.