Recently, I was working through some challenges with a custom schema that supports the structured content model for one of my pharma customers. I needed to take an existing document and lay it out with both the old schema and the new schema and make sure all the proposed changes would work.
I used a spreadsheet to help me organize my thoughts and see the content and the schema side-by-side. I started getting into a groove as I mapped the old schema to the new one. My entries into the “New” column started following a pattern:
- Content Type A
- Content Type B
- Content Type A
- Content Type B
At which point, Microsoft Excel automatically filled in the remaining 600 rows of the spreadsheet. Brrrrrrrzip! I could almost hear the wheels turning as the entries appeared.
Accept My Niceness
We have a phrase in my household: “ACCEPT MY NICENESS!”
We use this phrase when we realize that one of us is imposing help rather than offering help. Imposing help means helping without asking first, often for things as simple as:
- Flipping the pancakes that someone else is making
- Explaining how someone else should hold their knife
- Rushing over to take a heavy box out of someone else’s arms
This phrase must be spoken in all capital letters. Preferably, it should be slightly screeched, the way a pirate’s parrot companion might squawk, “pieces o’ eight!”
The point is to have a friendly and funny way to let each other know that (1) we appreciate the desire to help and (2) we need to obtain consent first. Consensual help feels good and supportive. Forced help feels critical and can end up preventing people from doing what they were doing, even if your intention was to help.
AI My Niceness
In our book The Personalization Paradox, Val Swisher and I describe what companies need to do to prepare content for automated content delivery. We write about how standardizing content in all five dimensions makes it possible for AI-based systems to deliver personalized experiences to customers on a massive scale. We also discuss how much personal data the AI system needs so that it can match the right content to the right person at the right time.
At this very moment, a lot of smart folks around the world are working out the guidelines for collecting, managing, and using this personal data ethically. Some of these guidelines have already made it into legislation, particularly in Europe and in California. The main theme so far? Consent.
We can make it easier for customers to find and use our information. We can help them have more relevant experiences. We can build “smart” software that looks for patterns and autocompletes, autofills, or auto-takes-over repetitive tasks.
But we’ve got to get consent, first. Otherwise we’re just imposing our help upon them while screaming, “ACCEPT MY NICENESS!”
Silence Cannot Equal Consent
If Excel had asked me before it filled in the remaining 600 rows of the spreadsheet, I’d have said “No thank you, not yet.” I had not yet fully figured out the new schema. I was only experimenting. I did not know if every part of the content model would map to the proposed new schema according to this pattern.
As soon as the software filled in for me, it changed my context. I had been working in the “create” part of my brain — organizing my thoughts, solving the problems, correcting my proposed new schema as I went along. Suddenly, I was pushed into “review” mode. I had to verify whether Excel’s guesses were true.
Once it was in front of me, all formatted and detailed and endless, it was difficult for me to find the errors. There were several areas where the schema needed to deviate from the pattern that had emerged in the early stages. Those areas were not obvious, once the entire column had filled in.
One of the key characteristics that makes something an AI and not just a collection of algorithms is the AI’s ability to learn on its own. We have an opportunity to teach it to ask first.
Silence cannot equal consent. We need to make it obvious for our customers to opt in, not just to our mailing lists but to our product features. Not just once, but each time, until they say “don’t ask me again.”
And if in fact, Excel did pop up some kind of “Want me to autofill?” message, then I didn’t see it. I was working pretty fast and hitting Enter pretty often. Any pop-up that appeared with “OK” pre-selected didn’t have a chance.
The truth is, our best attempts to make manual tasks easier, faster, and more accurate cannot succeed if we force our help upon people when they don’t want it.
P.S. As for the pancakes, well, for anyone who asks, mostly the key is not to over stir when you make them from scratch. Merely moisten the ingredients, even if the batter looks lumpy. Use medium heat instead of high. Flip ‘em when the little bubbles start to appear.
- 5 Tips for Selecting Your First Pharma SCA Use Case - January 24, 2023
- Reversing the risk: It’s time to adopt structured content authoring - September 19, 2022
- Single-source publishing to multiple outputs for Pharma - September 12, 2022