Online Learning: Google Apps... with AI?

Exploration of Google Apps... with AI?

For this design experiment, I explored several Google apps in a new context! I have been using Google apps for year, from Sheets, Docs and Slides to create and collaborate within documents to Maps and Translate when I travel to Sites and Blogger to share online content. While I've played around some with Google Classroom, my school doesn't use it, so I haven't been able to effectively test it with student populations or with a school framework.

However, early this week, I got a well-timed invite to help test a new version of some familiar Google tools using Google's own (still under development) in-house AI called Google Workspace Labs. 

Figure 1: Screenshot of M. Miller's "trusted tester" invitation for Google Workspace Labs

When I got this invitation I was intrigued! Only a limited number of "tester" invitations have been sent out, so this is a sneak peak at some software still in development! From what I could tell, Workspace Labs functions as an AI add-on for apps like Google Sheets, Google Docs and Google Slides, and integrates AI tools to accomplish tasks like organizing data in Sheets, creating customized images in Slides and writing content in Docs. 

I was ready to dive in and answer my questions about it: How well does it work? Could this be useful in the classroom or as a teacher preparing lessons? How might students use (or misuse) this new AI tool?

Diving In to Google's New AI

As this new "Duet AI" software from Google is being rolled out slowly to specific "testers", it is not well-supported or easily accessible. I wasn't able to find any useful tutorial yet for how to get started, although I'm guessing those will start popping up soon! While I wasn't surprised not to find any tutorial on YouTube yet, even the every-trendy TikTok didn't have much in the the way of advice for how to get started.

AI with Google Sheets

So I dove right in anyway! Using a small sample set of data, I started in Google Sheets and tried to get the new AI to organize and create some charts analyzing me data. (Something I figured was a pretty simple task that had taken me about 15 minutes to accomplish manually.)

Here were the results:

Video 1: M. Miller playtesting Google Workspace Labs AI in Google Sheets

As you can see, this was a pretty futile exercise. The current AI has very little understanding of how to translate common text commands into useful functions and, despite multiple attempts, I was never able to get it to create graphs. At most, it will create independent (and pretty!) tables, which may or may not reflect the actual data that you are trying to analyze. Overall, I was say this is very unhelpful.

As far as classroom applications: If I want to create blank table templates for my student to fill in while doing a lab, I guess it's helpful at going from a blank sheet to a sheet with labelled rows and columns. In this way it could save me some preparation time. 

However, due to it's inability to read and process real data, I would lean toward this AI tool being more harmful than helpful for students who are trying to learn about the integrity of information gathering for either science or math statistics.

AI with Google Docs

Secondly, I tried testing out the Google Workspace AI features in Google Docs. This too was met with limited success.

Image 2: M. Miller playtesting Google Workspace Lab AI in Google Docs

I again used my sample data from Sheets and asked Google Docs to analyze it for me. While the results started off a little promising -- yes, these are the results of a survey of employees at a company -- it fell apart pretty quickly from there. The survey was NOT about job satisfaction; the survey was out of 4 NOT 5 rating levels; the average response was 3.6 out of 4 NOT 4.5 out of 5; and the rest of the analysis is basically garbage. 

The clincher is the caveat at the end: "This is a creative writing aid, and is not intended to be factual." So, if we are looking for truth, then Google Workspace Lab's AI is NOT the answer. 

As a teaching or learning tool, I guess there are some creative writing applications where this is helpful. However, most writing I do needs to be clear and factual. While creative writing prompts are fun... I don't see this AI tool as something I would turn to frequently to provide content for my math or science classes. 

Similar to the Workspace Labs add-on in Google Sheets above, I think this content generated in Google Docs by the AI is misleading for students who are actually trying to learn how to analyze real-word data.

AI with Google Slides

Finally, I attempted to test the Google Workspace Labs AI in Google Slides. However, due to the slow rollout of the "trusted tester" program, I don't have access to this feature yet. Rather than give up, I decided to turn to a third-party AI add-on for Google Slides and test it instead.

Again, the results were not particularly impressive:

Video 2: M. Miller playtesting GPT AI in Google Sides

As you can see, I again attempted to generate an analysis of my data from Google Sheets, but this time in Slides. I liked how with this program you can select the number of slide and choose whether or not to add images! However the results were next to useless, yet again.

Generically, it generated an outline of what the titles of analysis slides might be. However, there was not actual analysis or real data, and the images, while pretty, had little context.

In regards to a teaching or learning aid, maybe this type of slide generator would be useful with a simpler prompt. However, for generating and sharing factual information with students, and specifically referenced data, this has little-to-no application in the classroom.

Conclusions

With regard to the TPACK model, Google Workspace Labs as an AI add on for Google Sheets and Docs (and probably Slides) is not there yet. 

T

The technology is accessible. Once granted access it's easy to integrate into existing learning tools. The use of common language prompts should theoretically increase accessibility even more, making things like blank pages and worksheets less intimidating. The limitations of the technology to only generate empty tables (but not manipulate real data) in Slides, only create creative writing prompts (but not anything factual or referenced) in Docs, and only generate images (but not content) in Sides, limits the current applications.

P

The pedagogy of having students use open-ended tools (like Docs, Sheets and Slides) to create as part of their practice and demonstration of mastery is consistent with the pedagogy of how I want to teach. As far as using creative tools, I don't want to hold my students back by not teaching them how to use common-language prompting tools like chat-based AI feature. However, teaching my students the ability to recognize the truth vs. lies and how to analyze real data is another essential part of my pedagogy. The extent to which this tool is misleading in it's misinformation makes it's applications limited to the pedagogical context of "today we're going to learn why we shouldn't trust AI", which while a valuable lesson, isn't always the direction I want to go. 

C/K

The primary objection I have to the use of the current Workspace Labs AI model is in the realm of content / knowledge. The recognition of truthfulness and the use of real, factual, referenced, reliable data and information upon which to make conclusion is essential for learning to take place. The current AI generates content, but without any regard as to its truthfulness. For this reason primarily, I would be reluctant to make much use of Google Workspace Labs AI in my classroom... yet!

This isn't to say that I am against using AI tools or that I think all AI is bad. This particular AI model is not developed to a point where it generates helpful content, yet. But maybe that's all the more reason to have students look at it now! If we wait until AI is more convincing at being trustworthy, maybe we won't develop the skills to question what's true. In the examples I generated above, it's ease for even the middle school students I teach to recognize that the AI is wrong and did not do a good job. If they see this now, maybe they will me more critical in years to come as the AI continues to get better at mimicking human tasks. Perhaps, in this context, a "broken" AI is a better teacher than one that works?