Fall 2023

The Unizin Application Development community, in collaboration with Colorado State University, is hosting an onsite hackathon on Thursday, October 26th (full day). Friday, 27th (½ day) in Fort Collins, CO. This event is designed to bring Unizin members together to gain hands-on experience working with Unizin and Unizin vendor partner products and services, such as:

• The UDP and associated data marts
• Working with Kaltura data
• Prototyping new application concepts (Canvancements, Denodo/Data Shipping)
• Expanding upon existing applications created by Unizin members

Unizin is working to have a synthetic data set for different activities, allowing all participants to directly access UDP data without privacy or security concerns. This event is primarily designed with application developers, business analysts, and data scientists in mind, though all Unizin members are welcome to attend.

Participants are welcome to stay at the Hilton Fort Collins, 425 W Prospect Rd. where we have a group rate of $149 per night. Use this link to secure the rate and room. Ten rooms are blocked at this rate until September 29. You can also call 970-482-2626 and ask for the Unizin Hackathon room block with booking code 90N. The other option is the Best Western University Inn which is a 5-10 minute walk to the TILT building where we will be meeting.

Denver International Airport is 65 miles from CSU. Groome Transportation is your best option for a shuttle directly to either Hotel. Book each leg separately since you’ll be leaving on Friday from the CSU Transit Center, a 5-minute walk from where we will meet.

From the airport: Select Hotels option: Hilton Fort Collins on Prospect

To the airport: Select schools option: CSU Transit Center

Flights on Friday after 4:15 PM are highly recommended and required by Groome for 12:05 pickup time. The scheduled dropoff for that shuttle is 2:05 p.m. If you have an earlier flight, you’ll need to select a 4:15 PM departure time when booking the shuttle. Shuttles are scheduled every hour and require a flight time 4 hours later. If you plan on attending, please let us know by adding your name to this registration form.

With Canvas Data 2 (CD2) implementation, Instructure is employing an entirely new data schema. This will do away with the dimension and fact tables The first focus for Unizin Engineering, Data Services, and Services teams was to successfully map CD2 to the Unizin Common Data Model (UCDM).  With minimal challenges and some remaining but small modifications to address foreign key relationships, Unizin has successfully made the translation. CD2 data should land successfully modeled within institutional members’ UDP tenants.  Queries and data marts generated through CD1 data will perform seamlessly with CD2 data.

So, what are the next steps?  What are Instructure’s timelines and how do they impact member institution’s transition?  And what about CD2 and the transition out of the UDW in Amazon Cloud?  

Over the next month, Unizin will finalize validation workflows and prepare to move CD2 data into member test environments.  Unizin has begun contacting members to schedule the transition  in October, November, and December.  This stage will focus on moving CD2 data to member SIT (System Integration Testing) environments, validating CD2 data against CD1 data in production, generating and conveying a validation report to members, and then switching CD2 to production. 

Through the SIT environments, both Unizin and member institutions will have working representations of CD2 data to validate existing code and workflows specific to each institution. However certain we believe we are matching CD2 to the UCDM, Unizin will verify the integrity of the data for each member before moving from their test environments to production.

“We intend to have CD2 migrations into member test environments and then again to production in place by the end of the calendar year,” said Bart Pursel, Unizin’s Chief Technology Officer.  “We are also actively engaged with Instructure in the event that we may need some additional time early in 2024 to complete the migration.  Currently, based on our mapping and preliminary work with the University of Michigan in testing CD2 in their SIT environment, we are confident we can implement CD2 for all members by the end of calendar year 2023.”

Another challenge operating in tandem with the CD2 migration is Unizin’s transition of the UDW out of Amazon Redshift and into Google Cloud (Big Query).  While distinct from the CD2  migration, the underlying change is both efficiency and the fact that the raw CD1 data is, of course, going away with CD2.  While members can expect a more efficient and adjacent repository of CD2 data in their Google Cloud Big Query project space, the schema has changed and will have some distinct differences in terms of scope.  For example, the Requests Table (now called “Web Logs”) \ is a massive table of system and user login data that will include only the most recent 30 days of data.  

For some members, the Web Logs table and the legacy UDW had little or no utility other than through usage in MyLA (My Learning Analytics).  MyLA previously depended upon the UDW.  Now that MyLA has fully transitioned \to rely on the UDP, the number of institutions requiring raw Canvas data to support workflows not utilizing modeled UDP data has decreased significantly.  Nonetheless, Unizin believes that providing an option for such data, especially if it is adjacent to modeled UDP data, will allow members to efficiently access raw Canvas data and make the adjustments to the changes Canvas has implemented with the least disruption possible.

There remain some additional unknowns, specifically around the possibilities of deltas or incremental data changes for CD2 data moving to the UDP, as well as what is both possible and practical in terms of the frequency by which the UDP can make an API request for “fresh” data.  According to Instructure, Unizin can expect no more than a four-hour latency for CD2 data.  For such data elements as “assignment_grades,” Unizin would like to eventually address specific tables and/or elements for targeted updates (with even less latency than four hours) without the overhead of a full data transfer. Regardless, Unizin will continue to work closely with members and Instructure to make this significant migration as seamless and transparent as possible.

Unizin Data Services and Solutions (DSS) completed their fifth DSS Town Hall on August the 8th of this year.  With over 50 in attendance, DSS updated members on a variety of Unizin and DSS-centered topics and projects, but with a twist this year: adding BeeKast survey and interactivity tool to provide real-time and asynchronous feedback on discussion topics.

The first request for feedback focused on privacy.  Unizin, either through security reviews or new member engagements, routinely addresses questions around its role in privacy, the UDP, and member institutional data. While Unizin has limited direct control over privacy policy and practices at member institutions, Unizin has yet to formally document its role in privacy in relation to member institutions. To address this gap, Unizin has drafted for member feedback a  “Unizin Privacy Statement” outlining Unizin’s role in confidentiality and security  (protecting data and the systems that house and process data), member roles in implementing privacy (protecting individuals) such as determining what data elements to convey to the UDP and transparency around the use of learning analytics. This privacy statement will remain “living” and open to modifications based on member input, but serve as Unizin’s and in general member’s roles and responsibilities governing privacy.

Beyond the privacy statement,  DSS shifted to the first real-time survey question exploring the formation of a privacy-focused working group.  Based on the variety of challenges and perspectives across the consortium and the impact these privacy positions have on the implementation and use of the UDP by member institutions, DSS wants to understand whether or not a new T&L group around privacy would be useful to members.

Would you be in favor of participating in or supporting the existence of a Unizin Working Group on Privacy, sharing common practices and challenges, and helping develop policy and/or practices that would facilitate the effective and ethical use of UDP data?

While many felt that they were not the best person at their institution to weigh in on such a group, the results were largely positive, with almost 70% at least “somewhat interested.”

The next steps will involve articulating the need for such a T&L group to the Unizin Board, developing a charter, and seeking membership.

The second Survey centered on Canvas Data 2 (CD2) and particular areas of concern to members.  While there are specific areas that must be addressed as priorities (mapping, validation, migration to production) the feedback we received helps us better prepare from a service delivery perspective to address particular pain points or areas of concern.  Areas of concern included:

The third and final survey question centered on our Synthetic data update. In the previous two DSS Town Halls, Sara discussed in some detail how DSS constructs synthetic or “fake” data derived from randomized distributions and census data to achieve representations of learner and learning environments.  Sara underscored that, while maintaining the rational shape of the data, synthetic data has no connection to an actual individual, course, or faculty member and no connection to anonymized data.  Synthetic data differs significantly from anonymized or de-personalized data in that it looks broadly at distributions of data and generates randomized, fictitious values based on generalized patterns found across all member institutions, for data such as course enrollments or grades, for example. Anonymized data involves simply the removal of identifiers from real data, preserving the integrity of the record while protecting the identity of the individual.

The synthetic data “word cloud” question allowed participants to candidly enter words or phrases in response to the prompt “What are some questions/concerns you have around synthetic UDP data?  Responses included:

• Much needed!

• Control of parameters

• Bias

• Algorithm for data construction

• Accuracy of representation of events

• Authenticity

• Ask my privacy, legal office

• Ability to represent a diversity

• Is it institution specific?

• Properly authorized by each institution

• Volume of data meets dashboard needs

• Non-identifiable institutions

• Did you already test its validity?

• Exclude inappropriate data

• When will it be available?

• Varied LMS usage patterns

• Representation of submission times

• Who needs to vett this?

The above feedback and associated comments that accompanied some of the responses underscore the need for data that can be used in multiple ways (safely introduce the UDP to new members or newly introduced existing members; training and institutional course work; proof of concepts for research or new implementations, etc.), and verification around methods used to generate synthetic data, and institutional review and validation of the data itself.

Across the last eight months, the Unizin integrations team (a collaborative effort between Engineering and Data Services and Solutions) has worked steadily over the past eight months with Packback to deliver Caliper-compliant events to the UDP.  Diligence and collaboration have led to success.  Packback provides a single resource for guiding and assessing writing assignments (including plagiarism detection), employing AI to provide students with immediate, meaningful feedback while they write. 

The new events leverage the Caliper Forum profile, providing metadata around students posting to Packback. The focus of these collaborations centers on key identifiers positioned in the correct format in the expected location of the event.  For example, it is critical to be able to identify from which course an event was launched.  Within Caliper, a “federated_session” block reflects the LTI launch from Canvas to Packback (or any third-party tool).  In this way, LMS IDs that the UDP uses to identify courses are passed through to the LTI application, and this pass-through should have consistent representation and positioning within the event. 

With data from a continually growing number of tools, members now have a new data stream that can easily be integrated into student activity analytics work, providing additional data that helps to more accurately represent individual student activity within a course, across a semester, or throughout an entire degree program. 

With Packback events now fully and consistently landing in the UDP, Unizin has begun announcing the availability of event data from Packback for those member institutions who have Packback adoptions at their institution. In addition to spotlighting this new tool at the August DSS Town Hall, Unizin will also reach out to institutional data stewards to inform them of this new application’s data moving to their respective UDP tenant.

Increasing the Diversity of Learning Tool Data in the UDP  (and working with vendors to configure and stream event data from their learning tools) is one of three Board priorities for 2023. Unizin’s work with third parties – particularly Packback and Ex Libris – to deliver consistently formed events is critical to this Board imperative and continues to enrich the learning tool data modeled and available for research through the UDP.

As discussed in our spring Unizin Report, Shane Nackerud and Kirsten Clark at the University of Minnesota, in addition to other member stakeholders and Unizin, have collaborated extensively with Ex Libris to deliver events generated through their Leganto (delivered through LTI in Canvas) reading list platform to the Unizin Data Platform (UDP). These events will reflect navigation to Leganto, reading list view, item access, and whether or not the content was downloaded. Employing the Caliper Library profile through Leganto moves us a step closer to incorporating resources originating from Library collections, providing a launchpad into what content students are viewing and potentially aligning these content actions with course contexts and behaviors broadly within the LMS.

The first run of test events was conveyed to the UDP in March, and DSS reviewed and conveyed changes around the needed location of Caliper objects and properly formatted identifiers.  The second set of events was sent in early July, with most of the required changes implemented.  Small changes around consistent identifiers that link events to Canvas courses are needed to complete fully formed Caliper events from Leganto – we are almost there, adding yet another tool’s unique representation of learner data to the UDP.

With new members, the UDP is exposed to additional member campuses.  New UDP implementations are underway. Unizin is excited to work with our new members and campuses on successful UDP implementations, as well as steward existing members that do not yet have a UDP, on different pathways forward to bring a UDP to life to quickly realize new insights from the learner and learning environment data.  

• UCI status – With a full SIS mapping all but complete, the University of California Irvine is poised to move from their SIT (System Integration Testing) environment to production.  Since late March, Unizin has worked with an amazing UCI team to map and align institutional SIS entities and elements to the Unizin Common Data Model (UCDM). In parallel, UCI has been developing a strategy for a learning analytics platform utilizing the UDP, and will “hit the ground running” once their UDP is migrated fully to production.
• Chapman – next steps – Unizin held an initial introduction and resources meeting with key Chapman stakeholders in early August.  Next, Unizin Services and DSS will initiate an onboarding meeting with Chapman that will look to implement Canvas context and event data as part of phase 1 of their UDP implementation.  Parallel to this first phase will be to implement the three or four high-value SIS entities, with additional entities added iteratively over the next year.

University of Michigan Flint update – with a primary focus on implementing My Learning Analytics (MyLA) at the University of Michigan Flint campus, collaborating with the Ann Arbor campus, the University of Michigan, Flint, is nearly ready to deploy its UDP into production.

Spring term billing activities wrapped up in April for late starting spring courses, which allows us to close out the books for the 2022-2023 academic year.  Unizin. We closed the year strong. Student Savings spring totaled $21.1M.  Unizin’s number of billable enrollments was 471K. This was a record high for the 22-23 academic year, as well as a record low for opt-out percentage, 0.58%!  *The graphic below also includes the summer term data).

On September 12th, Unizin released a Beta version of an Instructor Analytics Dashboard, that visualizes student reading events that take place in RedShelf, easily allowing instructors to better understand student reading behaviors. Before Unizin’s adoption of the RedShelf eReader, instructors had access to something similar in the retired Engage eReader. 

These visualizations can be used in multiple ways, such as helping instructors identify students who aren’t engaged in reading activities early in a semester, as well as using reading behavior data to inform what topics are important to emphasize in any given course meeting. The reading events that Unizin uses to power the dashboard are also available in a member’s Unizin Data Platform (UDP) event store, allowing for integration into larger analytics efforts to paint a more complete picture of student activity within a course. Unizin’s Resource Site contains information on how to best utilize the visualization.