Logging Interactive Visualizations
&
Visualizing Interaction Logs

This workshop takes place at the beginning of the upcoming IEEE VIS 2016 conference (23-28 October 2016, Baltimore, MD, USA), which is the premier venue for visualization. The workshop is open to anyone who previously registered at the conference. This website is still in progress, please come back frequently to get updates (and see recent changes here) or contact us directly.

Proceedings

Download the workshop proceedings

To cite this workshop:

Romain Vuillemot, Jeremy Boy, Aurélien Tabard, Charles Perin,
Jean-Daniel Fekete. LIVVIL: Logging Interactive Visualizations
and Visualizing Interaction Logs. Workshop IEEE VIS 2016,
Oct 2016, Baltimore, United States. 2016.

Or using BibTex:

@proceedings{vuillemot:hal-01535913,
 TITLE = ,
 AUTHOR = {Vuillemot, Romain and Boy, Jeremy and Tabard, Aur{\'e}lien and Perin, Charles and Fekete, Jean-Daniel},
 URL = {https://hal.inria.fr/hal-01535913},
 BOOKTITLE = ,
 ADDRESS = {Baltimore, United States},
 YEAR = {2016},
 MONTH = Oct,
 KEYWORDS = {Logs},
 PDF = {https://hal.inria.fr/hal-01535913/file/proceedings.pdf},
 HAL_ID = {hal-01535913},
 HAL_VERSION = {v1},
}
      

Keynote

"Visualizing and manipulating temporal event sequences with EventFlow: challenges and lessons learned for log analysis."

Speaker: Catherine Plaisant, Human-Computer Interaction Lab, University of Maryland.

Abstract: The Human-Computer Interaction Lab has developed tools such as EventFlow and Coco for event sequence analytics, which help analysts explore aggregated visual representations, search for specific patterns, compare groups of records, and interactively model the data to answer users’ specific questions. Through demonstrations I will quickly review the design elements that were most useful for log analysis and discuss challenges we encountered.

Bio: Catherine Plaisant is a Senior Research Scientist at the University of Maryland Institute for Advanced Computer Studies and Associate Director of Research of the Human-Computer Interaction Lab. Catherine Plaisant has over 140 published papers, on subjects as diverse as information visualization, medical informatics, universal access, digital humanities, technology for families, and evaluation methodologies. With Ben Shneiderman she co-authored the 4th, 5th and 6th Editions of Designing the User Interface, one of the major books on Human-Computer Interaction.

Schedule

Workshop Informations

This workshop addresses the following issues:

• How to log users interactions in a visualization system?
• How to design effective logging systems?
• How to make sense of those logs?
• How to deal with legal and ethic issues related to logging?
• How to design logs-based visualizations or any other application using logs data?

The workshop will be organized as follows:

• Position papers should express authors point of view; their length may range from a single paragraph, up to a 4-page article (using the IEEE TVCG format).
• Papers should be submitted by participants to romain.vuillemot@gmail.com and will be reviewed by the workshop organizers.
• Papers will be grouped by topic and presented by authors during panels, each will be introduced and moderated by an expert.
• One or many keynote researchers and practictionners--who heavily used logs in the past--will be invited to share their experience and present their vision on the main challenges.
• A concluding session will end the workshop to agree on main challenges and discuss next steps (journal submission, Dagstuhl seminar, ..).

The program may evolve depending on the number of submissions, but our goal is to allow as many people as possible to expose their point of view on this important topic. We invite IEEE Vis participants to contribute to this workshop in order to achieve the following goals:

Defining Logging Strategies

The first step when setting up a logging process is to ask what should be recorded, when it should be recorded, and how (by the web server, by the application itself, etc.). For example, even a simple and ubiquitous interaction, like a mouse dragging, requires carefully considerations as it can generate a lot of noisy events resulting in very large and thus difficult to interpret log files.

• How to record low-level interactions (mouse moves, keystroke, ...) and data-intensive interactions (dynamic queries, brushing and linking, ...) efficiently?
• How to track multiple and coordinated views? How to track the view the user currently focuses on?
• What is the scope of the context that should be recorded beyond user's interaction? Desktop UI configuration? Computer and office setup?
• How to record collaborative and multi-device activities?
• How to reduce the size of data intensive interactions? Should there by a low frequency / interaction sampling, filtering and/or aggregation to shrink log files? Should there be any buffering strategy?
• Is the Common Log Format (IP, User ID, Timestamp, etc.) generated by web servers, expressive enough to be the standard for Infovis? What are the related and upcoming standards (W3C, others)? Should Infovis define its own log format?

Logging Infrastructure

As we have mentioned earlier, a series of tools log users by default (e.g. proxies, web servers). However, from the authors' practical experience, it is oftentimes necessary to build its own tools for the sake of control over the logging format and flexibility in types of events to tracks.

• What is a simple and affordable setup for logging in Infovis?
• How to deal with offline tracking, synchronization? How to merge collected logs with other data sources, e.g. to clean, validate or enrich them with more contextual information?
• How updating an Infovis technique impacts previously collected/legacy logs?
• How existing APIs (Google Analytics, KissMetrics) can be used to track Infovis techniques? And perform tests such as A/B testing, perform cohort analysis, and real time monitoring?
• Beyond remote servers: what logging device or tracker can be used for logging? Can logging be manual and self-reported by users, instead of automated?

Making sense of logs: Analysis and Reporting

Logs reporting in academic papers varies with high discrepancies. This raises the need to improve logs analysis reporting to allow sound conclusions, and reproducibility of the evaluation.

• What relevance have vanity metrics (\# users, \# visits) to assess the success of an Infovis tool/technique?
• What should be the standard procedure or the best practices in logs reporting, for applications ranging from usability testing to evaluation?
• How to improve the reproducibility of research results and interoperability between logging tools and techniques?
• What are the specifics of logs for controlled experiments versus in the wild ones?
• How do user behavior framework like the HEART framework translate into logs? (and vice versa)

Legal and Ethical questions

As log collection and analysis is related to behavioral research involving humans, it requires approval from researchers' employer.

• How to make logging comply with IRB applications? How those applications shape the logging collection and evaluation procedure?
• What are the disclaimers best practices to notify users of logging activity?
• What would be the design of logging respecting privacy (e.g. logging that doesn't enable to reveal people's identity)?

Application Related to Logging

Finally, we think that agreeing upon a logging format and infrastructure, would have spillovers such as data interoperability and allow more applications building upon logs. Letting users visualize logs, whether it is their own or others, is a rich and promising area to identify patterns, insights of large logs collections. Logs may also enrich the user experience with enhanced history navigation, browsing and monitoring. More Infovis and Visual Analytics application already make sense of logs and we expect the workshop to discuss challenges (data type, user tasks, scalability, etc.) coming with those types of visualizations.

Workshop organizers

Romain Vuillemot: is an Assistant Professor of Computer Science at the Ecole Centrale de Lyon, Université de Lyon, France.

Jeremy Boy: is currently a data visualization specialist at the United Nations Global Pulse New York Lab, USA.

Aurélien Tabard: is an Assistant Professor of Computer Science at the Université Lyon 1, France.

Charles Perin: is a postdoc at the University of Calgary in the Innovis group.

Jean-Daniel Fekete: is a Research Director and the Aviz team leader at INRIA Saclay, France.