I believe that predictive text and auto-generated responses inhibit creative thinking, at least a little. To test out this theory, I (Laura) began an email chain that was comprised almost completely of the auto-generated responses provided by Outlook. The person I was emailing (Laurie) was also instructed to respond with the auto-generated responses whenever possible.
The initial email chain looked like this:
“Hey there,
This is a test for auto-complete responses.
Laura”
“I have received this email.”
“Great, thanks for letting me know!”
“No problem. Have a great day!”
“Thank you! You too!”
“Hi Laura, I guess that was the end of predictive text.”
The three suggestions given to me were:
“Ok, thanks.”, “Yes, it is.”, and “Indeed!”,
which I knew would end the conversation, so I didn’t bother responding. After talking to a couple of my friends who are avid users of auto-generated responses, I decided to create a new thread, which looked like this (see video to watch autofill in action):
“Hi Laurie,
I'm writing in regards to auto-complete suggestions. It has come to my attention that these suggestions are, indeed sourced from a pool of data, and not in fact based on what we write ourselves. I had earlier thought that maybe Outlook (plus, apparently Gmail, and probably others) learned our style and our words, but it seems as though that isn't yet the case. Maybe, in the future, the algorithms will become so smart that mail servers will be able to write emails for us as if they were really us, with readers being unable to distinguish the two. The ultimate Turing Test! But for now, we will stick with the generated responses that have already been created.
Ta ta for now!
Laura
PS our good friend Grace said that when she signs her emails, after writing "Grace" Gmail adds ‘and peace’ after it. Ha!”
I was trying to sound as unique as possible, to see what that would do to the auto-generated responses. Would it make them more interesting, or would it throw off the responses and not be able to come up with anything? The response I received was pretty generic:
“Hi Laura,
That is very interesting! I hope that you can find more interesting information about this topic. Enjoy the rest of your day and have a great rest of your week. Perhaps we can meet to talk about this in more detail.
Hope to see you soon.
Laurie”
But, the auto-generated suggestions I received for this email were:
“Schedule a meeting”, “Thank you! You too!”, “Looking forward to seeing you too!”,
which weren’t exactly the most appropriate responses for that email. Yes, she mentioned meeting sometime, but that really wasn’t the main point of the email. It made me wonder about how much of the content was being analyzed and how much was being ignored. What are the hot words that Outlook is looking out for and flagging as important?
After these fairly unsuccessful threads, I decided to go a different route altogether. I looked up “how to send an email” on Google and copied & pasted the most generic email I could find:
“To whom it may concern,
Thanks for getting back to me. I am writing in reference to predictive text. Thank you for your patience and cooperation. If you have any questions or concerns, don’t hesitate to let me know, and I look forward to hearing from you.
Cheers,
Laura”
Which yielded a better conversation just using the auto suggestions:
“Great! Thank you so much. Hope to see you soon.”
“You're welcome, and thank you!”
“When are you available for a meeting next week.?”
“Anytime works for me.”
“Okay, thank you!”
“Of course!”
“Perfect, thank you! Those dates work for me”
“Great! See you then!”
“Got it, thank you.”
But I still wasn’t satisfied with where this left the conversation. So I decided to get creative with my responses while Laurie kept on using the automatic options.
“Ah, It seems as though the auto-complete suggestions have ceased! It's too bad that the only options available pretty much end the conversation. Oh well, maybe if we correspond further, it will help generate new responses.
See you soon,
Laura”
“My thoughts exactly. Fingers crossed!”
“Yes, fingers crossed, indeed. It's interesting how the auto-responses always end the conversation. I wonder if that's intentional, or not. I wonder what kind of other responses I can prompt it with in my email. What do you think?”
“My thoughts exactly.”
To this, I was prompted with:
"My thoughts exactly!", "Yes, I agree.", and "Thank you!"
which basically was a repeat of what was sent to me. "Are the suggestions helpful?" No, they're not, and I gave my feedback to the pool of data, hoping to give it some helpful information to help it learn.
“I didn't like my auto-responses, they basically just repeated what you'd said. It seems like there are a pretty finite amount of auto-responses. How quickly do you think it'll learn what to say? Let's talk about it in more depth, shall we?”
“Sounds good to me.”
“Right....sure”
“Received, thank you.”
“Of course! Seems like responding to the auto-responses with other auto-responses really just doesn't work and will always end a conversation. I wonder if the auto-responses can teach people something (like, how to appropriately respond, for example, much like spell check and grammar check can teach you proper spelling/grammar), or if it just inhibits laziness. I'll keep you posted with my findings. Let's talk about this some more soon”
After this last one, we decided to stop using predictive text for a couple of emails, and started talking a little about predictive text in and of itself. But I really disliked these generated responses. "What do you think?" should be followed up by what someone actually thinks, not an AI generated response that a) has no thoughts and b) literally says you have no thoughts ("My thoughts exactly!" is kind of the most lame response to someone asking for your thoughts).
After some back and fourth about predictive text, 1984, and Harrison Bergeron, I received:
“Not sure I know that one. But I did intentionally not use the predictive text options in this email.”
Which prompted me with the responses of:
“Okay, thanks.”, “No, you did not.”, “Ok, thank you.”.
Two responses were basically identical, and the other I thought was kind of a cheeky response to not using predictive text, which was also the first time I’d seen that specific response. It was probably picking up on the fact that there were a couple of “not”s in the email. But it also made me wonder about what her predictive options were. We were talking about literature and intelligence, what would Outlook have suggested as a response?
Based on these email threads, auto-generated responses inhibit creative thought, and they almost always end the conversation. At this point in time, one can’t rely completely on auto-responses to hold a conversation. It's evident that when using the auto-generated responses back and fourth, the AI will get a bit lost, and will require more thoughtful responses to feed it data & ideas.
Explore how our words interact with each other on Voyant.
Stoplist words:
auto
complete
laura
laurie
predictive
reply
responding
response
responses
suggestion
text