Data Update No. 7
Data update No. 7 shares the JDD Lab’s analysis of Reddit data to July 2024. The Goal of the Justice Data and Design Lab’s (JDD Lab) is to use interdisciplinary teams of graduate students (currently Law and Data Science students) to collect and analyze person-centred data and user experience to improve British Columbian’s access to justice.
The JDD Lab is based at the Access to Justice Centre for Excellence (ACE) at the University of Victoria. Lab Director Kate Gower and a team of graduate students use machine learning and AI to collect and analyze new data – today’s data is from the social media site, Reddit1.
Here is what is going on:
1. New and Improved Clusters!
Our data scientists have been digging into the clusters in our data. We looked into the posts in last month’s tight cluster around insurance, and the posts on housing we gathered for the Alberta Law Reform Institute.
We discovered duplicates in the data. We identified two reasons: first, redditors sometimes re-post their queries, and second, our data-pull mechanism needed some updating. The JDD Interns took care of both things. We removed about 250 posts from the Reddit data lake, and then collected about 250 new posts from Reddit in July.
Result: New Clusters!
Cluster 5: Employment – This is the most cohesive cluster. The top terms are Employer, Employee, Hour, Working, and Employment. This cluster has the second most posts.
Cluster 4: *New: Contracts?* – This new cluster has the most posts, and the top terms are “contract”, “payment”, “business”, “email”, “canada” and “document”.
Explore the clusters below. Remember to set the Relevance slider to “0.6”2 for the best chance of identifying what the clusters are about (the Relevance slider is at the top right of the Report). We discussed Relevance in the LDA Report in Data Update No. 2.
Why this matters: This cleaning and recalibrating is the kind of iterative improvement which the JDD Lab is designed to do.
The JDD Lab shares its journey with you, so you learn how we get the best data we can offer. We think that a rising tide lifts all boats. In our case: a rise in data competence lifts all A2J enthusiasts :).
What we learned: The new cluster intrigues us. After the top six terms, it continues with Date, Received, Signed, Application and Letter. We are keen to figure out what help people are asking for in this cluster. Also, the small insurance cluster that appeared last month is gone, and we have a new top cluster regarding employment.
Data to July 2024 – Total Reddit Posts: 4362
| Cluster 5 Employment 1033 Posts | Employer | Employee | Hour | Working | Employment | Worked |
| Cluster 4 Contracts? 1437 Posts | Contract | Payment | Business | Canada | Document | |
| Cluster 3 Housing 853 Posts | Lease | Unit | Tenancy | Rental | Roommate | Apartment |
| Cluster 2 Family 340 Posts | Child | Parent | Mom | Mother | Dad | Father |
2. New Video – How to Prototype!
Research shows British Columbians look for legal advice online, and that most British Columbians look for help on the social media site, Reddit. The JDD Lab built a reddit bot – hence, “Rbot” – to meet people on reddit and offer help.
In Spring 2024, we collaborated with a team of graduate students at Vancouver’s Centre for Digital Media to build out the Rbot. They showed us the benefits of design-thinking and an agile, iterative production process. Watch this new video (run time: 32 minutes) to learn how the CDM prototypes.
We are grateful for the support of the Law Foundation of BC and Mitacs. We could not do this work without them.
- The most recent Everyday Legal Needs survey undertaken by Statistics Canada in 2021 shows that most people take action to resolve their everyday legal problems, and the top two things most people do are to ask their family and friends and to look on the internet. The JDD Lab used programs that show where people go online when they look for legal advice, and found that the top place people go is to the social media platform Reddit. ↩︎
- For more information on why, see Data Update No. 2, or see: Carson Sievert and Kenneth E. Shirley, “LDAvis: A method for visualizing and interpreting topics” (2014) Proceedings of the Workshop on Interactive Language Learning, Visualization, and Interfaces, Baltimore, Maryland, USA, June 27 at 67, online: The Stanford Natural Language Processing Group. ↩︎