Submission Deadline: June 15, 2023 (AoE)
The Student Challenge aims to foster new and nascent design ideas featuring around a specific challenge of relevance to the UbiComp/ISWC community this year.
Prompt Engineering for Behavioral Data
Are you eager to unlock more natural and engaging methods to understand passive sensing data collected from mobile devices? Join us in the UbiComp/ISWC Student Challenge, where teams of students from diverse academic backgrounds are invited to unleash their creativity and develop innovative approaches to exploring sensor data.
This challenge offers a unique platform to ideate, innovate, and showcase your ideas and creations to the UbiComp/ISWC conference attendees. This year, we specifically invite participants to create innovative solutions that empower users to ask questions and gain insights into their own sleep patterns. As part of the challenge, you will utilize a specific subset of the GLOBEM datasets that is specifically curated for sleep analysis.
The UbiComp/ISWC Student Challenge presents a remarkable opportunity for you to showcase your work and creativity to esteemed leaders in the field, and transform your innovative ideas into tangible reality. This year, a jury will review and evaluate entries, casting their votes for the Most Creative and Best Implementation categories, while conference attendees will have the opportunity to select their favorite teams for the People’s Choice award.
Summary of Key Dates
June 15, 2023: Submit your idea proposal (max 1 page) and register your team using this form.
June 30, 2023: Await notification regarding the acceptance of your submission.
July 1, 2023: Begin the development phase of the project.
October 1, 2023: Submit the short paper (max 4 pages), ethics statement (max 1 page), video demo (max 3 minutes), and code repository of your solution (optional).
October 7, 2023: Complete the development of the solution.
October 8, 2023: Participate in the conference and present a poster and live demonstration of your project
The GLOBEM Dataset
To participate, you should use a subset of the GLOBEM datasets, specifically focusing on the dataset named ‘INS-W_1’, which was collected in 2018 (prior to the COVID pandemic). The GLOBEM datasets contain multi-year mobile and wearable sensing datasets that include four years of data collection studies (2018-2021), conducted at University of Washington.
For the challenge, your focus will be on the sleep data collected within this dataset. You can find a detailed process for accessing the dataset by following this link. Additionally, you can find comprehensive information about all sleep-related features on this page.
For those seeking a greater challenge, you are welcome to explore the entire dataset, enabling you to delve into additional behavioral data beyond the focused subset.
Idea Submission (Deadline: June 15, 2023)
All participating teams must submit their ideas using this form. The submission deadline is June 15, 2023. It is encouraged that teams explore the ‘INS-W_1’ dataset from GLOBEM before submission.
Your submission should include the following:
Names and contact information for your team members: Provide names, institution, and email address of 2-5 team members, and contact information for the team leader.
Topic/Title of your idea: Provide the general topic, or a specific title for the idea you want to develop in your challenge.
Description of your idea (max 1 page): Provide a summary of your idea and how you intend to use the dataset.
Link to supporting materials (Optional): Include links to any relevant supporting materials such as images, videos, PowerPoint presentations, or other multimedia resources.
Completion of registration form: Ensure that all required fields on the registration form are filled out accurately, including contact information and any other necessary details.
Review and Notification (Deadline: June 30, 2023)
The student challenge chairs will review your idea submission and notify you by June 30 if your submission has been accepted.
Demo Development (July 1, 2023 to October 7, 2023)
The demo development phase starts after the notification of acceptance (July 1st, 2023) and lasts until 1 day before the presentation at the conference (October 7, 2023).
A technical kickoff session may be held around mid-July, to answer any questions you might have regarding the GLOBEM dataset.
Project Submission (Deadline: October 1st, 2023)
Each participating team should submit the following materials one week prior to their conference presentation:
Short paper (max 4 pages): The paper should include the motivation, background, and technical description of the solution, and include a minimum of 4 references. Please use the UbiComp/ISWC templates and guidelines to prepare the short paper:
Ethics Statement (max 1 page): The statement should provide a clear description of how the technology benefits users, including potential malicious uses beyond privacy concerns. Additionally, present a vision for an engineering or hybrid solution to mitigate or eliminate harm associated with the technology, considering ethical, social, and security implications.
Video Demo (max 3 minutes): The video should showcase a complete use case of the technology, and incorporate storytelling elements to provide context and enhance understanding of the submission.
Code repository (Optional): While not mandatory, we encourage you to open-source your code, as it allows for greater transparency and collaboration. If you do, please include clear instructions for installation and usage to facilitate evaluation and replication of your solution. It is important to note that the quality of the code will not be a factor in the judging process, so feel free to share your code regardless of its current state.
Teams will showcase their demos at UbiComp/ISWC 2023. Alongside the demo presentation, please prepare a poster that concisely summarizes your idea, design, and implementation. Evaluations will be conducted by a jury and the audience, with the chance to win cash prizes in categories such as Most Creative, Best Implementation, and People’s Choice Awards.
Zero-shot LLM: Explore ways to make large language models understand sensor data and enable users to answer sleep-related questions.
Example Prompt: Given step count data, how many hours of deep sleep did I have last night?
Data-driven Storytelling: Generate interpretations and insights from user data by employing techniques such as correlation analysis, clustering, or causal inference. Large language models can also be particularly useful in this context.
Example Prompt: Given phone usage data, why is my sleep quality trending up/down?
Sensor-based Prediction: Leverage data from additional sensors to enhance sleep outcome predictions. By combining sleep data history with other sensor modalities, develop models that can predict sleep outcomes.
Input: Sleep data history, other sensor modality
Output: Sleep outcome
Competition Rules, Details, and Suggestions
Team Size and Composition: We require teams to be composed of 2 to 5 students. The students are not required to be from the same institution and can come from different educational backgrounds (Bachelors, Masters, and PhDs).
Number of Teams to be Selected: Depending on the number of teams, we will select between 40-50% of the teams, with a maximum of 18 teams presenting at the conference.
Judging Criteria: We will judge projects based on four categories: creativity (how novel the idea is, how it combines multiple ideas from different disciplines), engineering (how strong is the demo implementation), ethics (we require an Ethics Statement specifically added to the submission), impact & scalability (how likely would it be that this technology can be implemented in real life and will touch many lives).
Registration Requirement: Please register for the contest and submit your idea using the form. If your team is accepted, at least one of your team members MUST register and attend the conference (or be accepted as a student volunteer). You will then be able to demonstrate your idea to the live audience, claim your prizes, and enjoy the awesome UbiComp/ISWC conference!
Demo Ideas: Teams will be permitted to demo only one idea; different demos of the same idea are permitted.
Demo Code: We encourage you to open-source your code (if you do, you can provide a link to your open source repo during your presentation). Note that teams will not be judged according to the quality of the code.
Resources for Development: Obtaining an API key from OpenAI or your preferred source, such as other language model providers, is highly recommended if you intend to utilize their GPT-models. For faster computation, consider GPU resources as well.
Extra Equipment: Teams will be required to bring all the hardware devices needed to present the demos to the conference. No additional equipment will be provided at UbiComp.
WiFi: WiFi is available at the conference; however, please be aware of bandwidth restrictions. Teams are welcome to use the conference WiFi as a part of their demos; however, we recommend having a Plan B, such as your own local network using routers.
Anonymity: Submissions are not anonymous and should include your team members’ names, affiliations, and a team/project name.
Submissions will be acknowledged and awarded in three distinct categories:
- Most Creative Work (Evaluated by the Jury)
- Best Implementation (Evaluated by the Jury)
- People’s Choice (Voted by Conference attendees)
For questions regarding to the challenge, please reach out to the Student Challenge Chairs at firstname.lastname@example.org
Student Challenge Chairs: Pablo Paredes (UMD), Jean Costa (TRI), Aleksandar Matic (Koa Health)