Before getting into this experiment I want to acknowledge that there is no magic “AI Button” – AI can do lots of things, so to define my terms a bit better, what I was interested in was whether using AI voice to text functionality as part of a reflective journal would improve the quality of reflection. The experiment is limited as the sample size of 1 (me), but I thought it would be useful learning about the reflective process as well as allowing me to hack together an MVP for actually doing it. As the MVP aspect was also important to me, I decided to use everyday tools that were likely available to a broad spectrum of people.

Hypothesis:

AI voice to text recognition will improve the quality of reflection when used with a reflective journal structure.

Process:

1. I created a simple reflective journal structure based around my work on worldviews. This was based around 4 key questions:

• What does it mean to be human?
• What is real and how do I know it?
• How is knowledge created and how do I know it is accurate?
• What is more important and how does it affect choices?

2. I set it out with text boxes in Powerpoint (desktop) and used the “Dictate” function. This was really frustrating and clunky and stopped working after about 100 words. After some googling I found out that Powerpoint in 365 works differently so I put the reflective journal structure into that and repeated the process. This worked really well, but I found that I had to do a lot of editing when the word count got high e.g. resizing the text box, and reformatting the journal so that there was only one response space per slide. I concluded that Word was a better tool for this. I repeated this step in Word (365), and was really pleased with the results. I was able to talk naturally and Word (365) accurately captured the words and inserted punctuation. Very little editing was needed after I read back my reflection. However I noticed that it was quite lengthy, I repeated myself and it wasn’t easy to grasp the key themes or points that I’d identified through the reflection.

3. I saved the document and uploaded it to ChatGPT, using the prompt “summarise the user’s perspective”. The results are impressive – these few extra steps had allowed me the freedom to reflect “free flow” with pauses, circular thinking and repitition, and output something that I could quickly and easily refer to, and (if needed) submit to someone in a format that they could easily understand my thought process. In short, writing was no longer a barrier to the quality of my reflections.

Output:

Here’s an excerpt from the reflective journal to illustrate my point about how verbose and circular I was in my reflections (I found the freedom to do this really helpful)

The longest part of this experiment was trying to do the same thing (voice to text) in many different everyday tools. In addition to Powerpoint and Word, I also tried Excel, Google Docs and Google Forms – in terms of ease of functionality and quality of voice to text, my preferred tool was Word (365). I did a full reflection in Ppt (365), and only responded to one question in Word, so the document that I uploaded to ChatGPT was Ppt.

Here is the summary that ChatGPT made of the complete (Ppt) reflective journal entry:

1. Being Human:
   • To be human involves self-doubt, growth, learning, and striving to be better.
   • It is about relationships, responsibilities to others, and living with diverse perspectives and expectations.
   • There’s a recognition of interconnectedness among humans and other life forms and a responsibility to leave a positive legacy for future generations.
2. Reality and Perception:
   • Reality is experienced through senses, feelings, and stories—both personal and shared.
   • Reality is co-constructed through dialogue and shared perspectives, emphasizing the exchange of ideas and insights for deeper understanding.
3. Knowledge and Accuracy:
   • Knowledge stems from experience, reflection, and ethical, transparent processes.
   • Its validity is emphasized over strict accuracy, with robust methodologies ensuring its relevance and trustworthiness.
4. Values and Choices:
   • Personal values include a love for novelty, challenge, and meaningful work over material gain or status symbols.
   • However, choices are shaped not only by values but also by lived experiences of assimilation, societal expectations, and historical family struggles, which often obscure or conflict with true values.

This perspective reflects a thoughtful, interconnected, and reflective approach to understanding self, others, and the broader world.

Was it useful / what did I learn?

For me personally, and professionally, this was a really useful exercise. I often create learning journals and reflective journals as part of development interventions, but completion is generally low (unless it is a requirement for a completion certificate) and the quality of engagement is also often low. The main reason cited for this is lack of time, difficulty to articulate and arrange thoughts, and sometimes the requirement of doing this in English can be a barrier. In terms of how AI can act as a personal assistant for reflection, using tools that are already present in frequently used software, being able to reflect aloud, but submit in text seems to reduce a lot of barriers. However does it actually improve reflection? I can’t generalise based on a sample size of 1 – however, I liked having this option. I found that I was able to just start talking, without feeling the need to plan or organise my thoughts, which was freeing. I didn’t have to worry about editing or getting my expression right, but I did have to get over my worry about sounding silly talking to my computer.

In terms of the implications, what I take away from this experiment is that embedding voice to text functionality in a reflective journal could reduce some of the barriers to engagement and quality of expression, but that the quality of reflection is still dependent on the individual choosing to go through the process. I was also excited to see that from an MVP perspective I could trial this more broadly with people, using the technologies and AI tools, that I already have. If I was doing this with a client, however, I would use CoPilot, rather than ChatGPT, and would have to assess any data privacy or commercial risks around doing this.

A next step for me is to see how AI could be used to analyse the outputs of reflective journals, as these can be a source of input for further coaching conversations as well as indicating the value that users are getting from a broader development intervention.