Link: https://blogs.ubc.ca/ideamachinery/2021/03/25/task-11-algorithms-of-predictive-text/

I selected to link to a Dierdre’s post on using a predictive text algorithm partially because this is one of the optional tasks that I selected not to complete my own posting on. The main reason for this was that I had already completed a sufficient number of optional tasks, and not due to some larger preference to avoid it. As such, I was curious to see what the task was all about. Admittedly, I hadn’t completed the week’s lesson prior to making my way through her post. As such, I was very appreciative of the detail that she went into in her post. Instead of simply reflecting on her “dude bro” predictive text and her feeling surrounding it, she provided a lot of background information on the algorithms behind the functionality. Honestly, I am quite impressed that she was able to navigate her predictive text to the point that it not only made sense, but came off as somewhat profound. Spurred on by her experience, I attempted my own predictive text message using the same given prompt “As a society, we are…” and was highly disappointed with the result, which ended up including my favourite author’s name (showing me the personalized nature of the algorithm), though with a misspelling which I must have accidentally used enough times that my G-mail phone app now recognizes the error as being correct.

Dierdre referenced Gmail’s Smart Compose functionality, which made me think of how when composing e-mails using Gmail on my computer (rather than on my phone), the autocomplete suggestions are much more robust and intelligible, often suggested in longer phrases rather than single words. I decided to attempt the same task using my desktop Gmail and found that I wasn’t actually provided with any prompts for the input. I think the reason for such is that it was too off-script from everyday conversational language. Prompts like “Hello, how…” will easily be completed with satisfactory suggestions, but more complex or unique prompts like the one Dierdre and I both selected stray too far from the basic conversational fillers that Smart Compose was designed to provide.

I actually wrote a speculative narrative that involves a subject working to create a database of predictive text phrases that allow users of these technologies to sound less robotic and more like their natural speaking styles while utilizes the technology: Alternate future #2: PRESCRIPTIVE TEXT.