What Does Predictive Text Spell for Human Writing?

Hello, my name is a very merry birthday. You killed my dog. Prepare to be the best burgers.

I think we should all do that tomorrow and I can take a look at the schedule.

It smells like you are going on the same day.

I was born to the internet in the internet for a few years and then it started with the same thing.

While you may not have been “born to the internet”, if you have spent any time with it in recent years you might recognize the italicized text as prompts used during popular predictive text “games.” (Try them for yourself! Simply type in the first few words and then let the autofill complete the rest of the sentence.)

As you may have guessed, this month’s Dataspace newsletter is all about predictive text – what it does and what this particular human/computer collaboration could spell (literally and figuratively) for communication.

Just the basics:
Check out these resources for an intro to predictive text in three minutes as well as a further discussion on some of the technical methodology. 

Beyond the tech specs of how these algorithms actually work, the growing capabilities of predictive text beg a multitude of questions regarding the potential implications for humans:

How is predictive text affecting how humans write?
Are these tools having a permanent effect on our written communication? Will certain terminology suddenly become more commonplace because it is continually reinforced by the predictive suggestions? Are we risking sacrificing thoughtful communication in the name of efficiency? At what point will the reality become that humans are not writing at all, merely selecting from a variety of computer generated suggestions? How will authors maintain a unique written voice? How does this technology change our relationship with our own thoughts? 

As one of the above articles points out, the predictive text suggested by the algorithms tend to be – not surprisingly – more predictable, and less creative. At the same time, we are more reliant than even on written communication (email, texting, etc.), and have been trending towards the habit of using emojis and emoticons to convey tone and subtext. If our text itself is bland because it has been formulated a bit more by the algorithm, we compensate for this by spicing it up with a couple of emojis, and then emphasize the point by following up with a gif.

Do these trends in communication signal that we could be in danger of losing the ability to fully utilize the richness of our actual languages? Will we start dropping the use of colorful adjectives and emotive words and instead rely more on visuals to convey this subtext?

Specifically, how is the technology of predictive text having an effect on children’s language learning?
There are valid concerns that have been raised about how the ability to “write” a sentence with just a tap of the finger isn’t doing young learners any favors. While the author of the linked article seems to have a bit of a chip on his shoulder regarding the education level of the youth in his country (Ireland) at this juncture, to some extent he does have a point. Repetition is an important part of the learning process. Much in the same way athletes perform repetitive drills and exercises to train their muscles to fire more rapidly in particular familiar ways, the repetition of mental tasks develops that mental “muscle memory.” In order for a child (or adult) to really learn how to spell a word and use it properly, they have to actually engage in the process of spelling and using the word – writing it out letter by letter over and over again.

The prevalence of predictive text in email, google docs, email, etc means that students could theoretically complete their entire education without having to actually spell the word “theoretically” on their own. The predictive text anticipates what they are trying to write and fills in the word for them, and autocorrect takes care of any errors before they have to run a spell check and manually check for any misspellings. (The Google Docs AI completed my words and corrected my spelling multiple times throughout the course of writing this paragraph.)

Looking at this question from another perspective, it is interesting to note that certain educational theories have recently come under fire for their methods of teaching children to read through a process of making educated guesses about what a word might be – theorizing that they can “predict” a word based on certain cues before they actually read it. The processes that are seen as potentially detrimental to the young readers’ learning are strikingly similar to the functionings of predictive text algorithms. These seem to reiterate the point that when humans focus too narrowly on speed and efficiency  – including in areas such as language learning and communication – we risk sacrificing other skills along the way.

Will we rely on AI to identify the writing of AI?
Will we be able to identify when an email, news article, text, dating profile etc. has been written by an AI instead of an actual human?  Will AI become proficient enough to start generating an infinite stream of “creative content”?

How strong will AI’s “understanding” of our language grow?
Will it become strong enough that it can provide valid advice into decoding our human-human interactions?

The above list of points worth pondering is certainly not exhaustive, and I am certain there are multiple questions and issues that we as humans have yet to explore as our relationship with predictive text continues to develop.

In the meantime, perhaps it is worth posing some of the above questions to the AI algorithms themselves?? Let’s see what the predictive text has to say…

AI will change learning by a couple weeks after work on the schedule.

Predictive text will not work on the phone.

Autofill will never be able to do it.

AI’s understanding of our language will be how it goes to the sunshine in a way.

Very insightful.

Until next time..
Cheers, everyone!

-Katie

BEN’S TAKE
While I am putting time into Golden Record (see below), Dataspace’s primary focus remains analytics and data engineering staffing and recruiting. As you probably know, our niche is in serving managers who just aren’t getting the quality of staff they need through their normal staffing providers. I think that message is resonating as we’re seeing an uptick in interest lately. If you’ve got a staffing or recruiting need, please do let me know. I’d love to see if we can help.

Also, I continue to receive an increasing number of compliments on these newsletters (sadly, for my ego, this trend coincided with my handing off much of the writing to Katie, our lead recruiter). If you know someone who might benefit from our perspective, please shoot me a note and I’ll add them to our email list.

GOLDEN RECORD
Golden Record, our technology for linking records across databases and data sets, is improving all the time. Here are some of the things we implemented in December…

  • We added the ability to see all keys in all related records when you query the golden set. In other words, with one query you can now find a record in any of your data sets along with the keys of the related records in all your other data sets.
  • We improved our logging capabilities to enable better customer support.
  • We also implemented version one of our API! Authorized developers can now access Golden Record via our RESTful API. The first cut of the API documentation is available here. (yes, the documentation needs some work). You’ll need an auth code to actually try out the calls. Let me know if you’d like me to set you up – I’d really appreciate the input!

This month we’re implementing a piece of functionality that will improve the quality of the matches we find. We’re also pursuing some fun marketing efforts. More on all of this next month. In the meantime, let me know if you’d like to know more about Golden Record and kick around your use cases.

You can reach me at Benjamin.Taub@Dataspace.com
That’s all for now. Thanks for reading and stay safe!
– Ben

JUST IN CASE YOU MISSED IT…

We’ve posted several new articles to our Dataspace Blog since our last newsletter. If you missed them the first time around, here’s your chance to check ’em out!

ALL I WANT FOR CHRISTMAS IS ANALYTICS
POLLSTERS SHOULD TAKE A HINT FROM THE CIA
ASSIMILATE.. OR BE ELIMINATED?
DATA PLANS FOR TURKEY DAY
DONUTS WILL FIX IT!
IS AI COMING FOR YOUR JOB?

SUGGESTIONS?

What do you think? How can we make this blog more useful to you? What topics would you like to see more of? Want to contribute an article? Just want to catch up and chat?

I’d love to hear from you! Email me at benjamin.taub@dataspace.com.
Thanks for reading!

Ben

0 replies

Leave a Reply

Want to join the discussion?
Feel free to contribute!

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Want more like this? Sign up for our newsletter!

0 + 5 = ?