Detecting and correcting user activity switches: algorithms and interfaces
Abstract
The TaskTracer system allows knowledge workers to define a set of activities that characterize their desktop work. It then associates with each user-defined activity the set of resources that the user accesses when performing that activity. In order to correctly associate resources with activities and provide useful activity-related services to the user, the system needs to know the current activity of the user at all times. It is often convenient for the user to explicitly declare which activity he/she is working on. But frequently the user forgets to do this. TaskTracer applies machine learning methods to detect undeclared activity switches and predict the correct activity of the user. This paper presents TaskPredictor2, a complete redesign of the activity predictor in TaskTracer and its notification user interface. TaskPredictor2 applies a novel online learning algorithm that is able to incorporate a richer set of features than our previous predictors. We prove an error bound for the algorithm and present experimental results that show improved accuracy and a 180-fold speedup on real user data. The user interface supports negotiated interruption and makes it easy for the user to correct both the predicted time of the task switch and the predicted activity.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
Previous automatic task switch detection approaches have a high deviation in time of 3 to 5 minutes between predicted and actual task switches.
“or that one has to accept a high deviation in time of 3 to 5 minutes between predicted and actual task switches [13], [14], [18].”
From Detecting Developers' Task Switches and Types (2022) · cites this paper as [14]
Previous approaches predominantly used longer and fixed window lengths of 5 or 10 minutes.
“In contrast, previous approaches predominantly used longer and fixed window lengths of 5 or 10 minutes [13], [14], [17], [27].”
From Detecting Developers' Task Switches and Types (2022) · cites this paper as [14]
A task switch is less likely immediately after a task switch, and the likelihood of a task switch increases as time passes.
“One feature captures the difference in the time spent in an application, since this might be another indicator for a task switch, either because a switch is less likely immediately after a task switch, and the likelihood of a task switch increases as time passes [14].”
From Detecting Developers' Task Switches and Types (2022) · cites this paper as [14]
Shen et al.'s approach is mainly based on semantic features and supervised learning.
“One of the most prominent approaches is by Shen et al. [13], [14], [16], [18] that is mainly based on semantic features and supervised learning.”
From Detecting Developers' Task Switches and Types (2022) · cites this paper as [14]
Shen et al.'s evaluation is based on a small set of two users and counts a task switch as accurate if it falls within a 4 to 5 minute time window of a real switch.
“Their evaluation is based on a small set of two users and counts a task switch as accurate if it falls within a 4 to 5 minute time window of a real switch”
From Detecting Developers' Task Switches and Types (2022) · cites this paper as [14]
The prediction delay of the authors' approach is considerably smaller than that of previous approaches that applied fixed window lengths of usually 5 minutes.
“This delay is considerably smaller compared to previous approaches that applied fixed window lengths of (usually) 5 minutes (e.g., [13], [14], [17], [27], [32], [33]).”
From Detecting Developers' Task Switches and Types (2022) · cites this paper as [14]
Cites
LIBSVM: A library for support vector machines
Chang, C.-C., & Lin, C.-J. (2011) · ACM Transactions on Intelligent Systems and Technology · doi:10.1145/1961189.1961199 · Cited by 40,918 · see citing works
The perceptron: A probabilistic model for information storage and organization in the brain.
Rosenblatt, F. (1958) · Psychological Review · doi:10.1037/h0042519 · Cited by 12,280 · see citing works
Libsvm : A library for support vector machines
Ferrari, V. (2008) · Medical Entomology and Zoology · Cited by 10,124 · see citing works
Online Passive-Aggressive Algorithms
Crammer, K., Dekel, O., Keshet, J., Shalev‐Shwartz, S., & Singer, Y. (2003) · Cited by 1,698 · see citing works
Learning to Classify Text Using Support Vector Machines
Joachims, T. (2002) · Springer eBooks · doi:10.1007/978-1-4615-0907-3 · Cited by 1,379 · see citing works
Learning to Classify Text Using Support Vector Machines: Methods, Theory and Algorithms
Joachims, T. (2002) · Cited by 933 · see citing works
Online Passive-Aggressive Algorithms
CrammerKoby, DekelOfer, KeshetJoseph, Shalev-ShwartzShai, & SingerYoram (2006) · Journal of Machine Learning Research · doi:10.5555/1248547.1248566 · Cited by 914 · see citing works
A diary study of task switching and interruptions
Czerwinski, M., Horvitz, E., & Wilhite, S. (2004) · Proceedings of the SIGCHI Conference on Human Factors in Computing Systems · doi:10.1145/985692.985715 · Cited by 830 · see citing works
Bellotti, V. M. E., Ducheneaut, N., Howard, M., & Smith, I. (2003) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/642611.642672 · Cited by 370 · see citing works
TaskTracer: a desktop environment to support multi-tasking knowledge workers
Dragunov, A. N., Dietterich, T. G., Johnsrude, K., McLaughlin, M., Li, L., & Herlocker, J. L. (2005) · Proceedings of the 10th international conference on Intelligent user interfaces · doi:10.1145/1040830.1040855 · Cited by 240 · see citing works
Self-citationKaptelinin, V. (2003) · ACM SIGCHI Conference on Human Factors in Computing Systems (CHI) · doi:10.1145/642611.642673 · Cited by 213 · see citing works
(2001) · Annals of Oncology · doi:10.1093/oxfordjournals.annonc.a000307 · Cited by 175 · see citing works
A hybrid learning system for recognizing user tasks from desktop activities and email messages
Shen, J., Li, L., Dietterich, T. G., & Herlocker, J. L. (2006) · Proceedings of the 11th international conference on Intelligent user interfaces · doi:10.1145/1111449.1111473 · Cited by 109 · see citing works
Self-citationSmith, G., Baudisch, P., Robertson, G. G., Czerwinski, M. P., Meyers, B. R., ROBBINS, D. W., & Andrews, D. E. (2003) · Cited by 98 · see citing works
The perception: a probabilistic model for information storage and organization in the brain
Rosenblatt, F. F. (1988) · Cited by 78 · see citing works
Matching attentional draw with utility in interruption
Gluck, J., Bunt, A., & McGrenere, J. (2007) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/1240624.1240631 · Cited by 70 · see citing works
Activity Explorer: Activity-centric collaboration from research to product
Geyer, W., Muller, M. J., Moore, M., Wilcox, E., Cheng, L.-T., Brownholtz, B., Hill, C. L., & Millen, D. R. (2006) · IBM Systems Journal · doi:10.1147/sj.454.0713 · Cited by 42 · see citing works
Fewer clicks and less frustration
Bao, X., Herlocker, J. L., & Dietterich, T. G. (2006) · ACM International Conference on Intelligent User Interfaces (IUI) · doi:10.1145/1111449.1111490 · Cited by 40 · see citing works
Self-citationReal-time detection of task switches of desktop users
Shen, J., Li, L., & Dietterich, T. G. (2007) · Cited by 28 · see citing works
Self-citation
Flagged first, then by citations
Cited by
Sorted by citations
Detecting Developers' Task Switches and Types
Meyer, A. N., Satterfield, C., Zuger, M., Kevic, K., Murphy, G. C., Zimmermann, T., & Fritz, T. (2022) · IEEE Transactions on Software Engineering · doi:10.1109/tse.2020.2984086 · Cited by 14 · see citing works
“Time spent per application category (13): the percentage of the total duration of the task segment that was spent in each of the 13 application categories [14], [32], [33] NA 34.”
BackgroundMethodResultInfluentialSwitch detector: an activity spotting system for desktop
Mirza, H. T., Chen, L., Chen, G., Hussain, I., & He, X. (2011) · Proceedings of the 20th ACM international conference on Information and knowledge management · doi:10.1145/2063576.2063947 · Cited by 5 · see citing works
- Authors
- Shen, J., Irvine, J., Bao, X., Goodman, M., Kolibaba, S., Tran, A., Carl, F., Kirschner, B., Stumpf, S., & Dietterich, T. G.
- Affiliations
- Oregon State University (US)
- Venue
- Proceedings of the 14th international conference on Intelligent user interfaces
- Publisher
- ACM
- Published
- 2009-02-08
- Pages
- 117-126
- OpenAlex
- W2079972412
- Cited by
- 37
- References
- 19
- Citations by year
- 2026: 1 · 2024: 1 · 2020: 2 · 2019: 3 · 2017: 2 · 2016: 3 · 2014: 1 · 2013: 3 · 2012: 4
- FWCI
- 5.71
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- ACM International Conference on Intelligent User Interfaces (IUI) (publishedVersion)
- ENLIGHTEN (Jurnal Bimbingan dan Konseling Islam) (submittedVersion)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Context-Aware Activity Recognition Systems
- Social Sciences › Decision Sciences › Information Systems and Management › Personal Information Management and User Behavior
- Physical Sciences › Computer Science › Human-Computer Interaction › Usability and User Interface Design
- Keywords
- activity prediction, online learning, error bounds, knowledge workers, user activity modeling
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
