Explained via cheese
Clearing the air around whether Grice’s maxims are even useful to conversation designers
What do fried Brussels sprouts, mom jeans, and Grice’s maxims have in common?
Against all odds, the three have managed to claw their way out of the depths of irrelevance and make a COMEBACK, baby.
*reviews notes* So, are we making a judgement call about this, boss?
I’ll leave the Brussels and jeans up to your own judgement. However, the maxims do need to be addressed.
If you have been hanging out lately in any conversation design circles, you’ve probably heard somewhere that your design should “follow Grice’s maxims”. Let’s…uh, unpack that.
In different cultures there are rules and norms that help people understand each other, and some of these rules regulate how conversation should be conducted.
One set of rules (in English) consists of the four conversational maxims, put forth by the linguist Paul Grice.
The four maxims are:
- The maxim of quantity, where one tries to be as informative as one possibly can, and gives as much information as is needed, and no more.
- The maxim of quality, where one tries to be truthful, and does not give information that is false or that is not supported by evidence.
- The maxim of relation, where one tries to be relevant, and says things that are pertinent to the discussion.
- The maxim of manner, when one tries to be as clear, as brief, and as orderly as one can in what one says, and where one avoids obscurity and ambiguity.
And one other rule, oft-quoted in the world of conversation design, is the cooperative principle: the idea that it is safe to assume that whenever you’re conversing with someone, you’re both doing your best to cooperate to achieve mutual intelligibility. Mutual intelligibility is a fancy way of saying that you both catch each other’s drift.
But like I said, if you’ve attended a voice tech conference in the last 3 years you probably know all this so far.
Unfortunately, right about here is where most of us stop, when we say things like “it’s important to keep Grice’s maxims in mind when designing usable and delightful voice user experiences.” But that’s a shame, because this is actually where all the fun begins. 🙂
These maxims and this principle sound nice in theory…but we all know that sometimes people say things they don’t mean. Or share waaay too much information on a first date. Or take way too long to tell a story (my mom is the queen of this).
Yeah, I really enjoyed hiking 16 miles through Yosemite on my birthday against my will.
Yeah, I really did NOT enjoy hiking 16 miles through Yosemite on my birthday against my will.
So, if it turns out that humans sometimes aren’t cooperative, relevant, concise, truthful, and orderly when conversing, doesn’t that mean that Grice was just totally off base with his maxims and his cooperative principle?
And if this is the case, how have we all not devolved into complete lingual chaos?
👁 👄 👁
Luckily, many of us linguists have spent countless hours and a boatload of cash on an education that helps answer this question. So for the approximately 34 people on Earth that care to know, we got you!
We all do stuff with Grice’s maxims
Grice is actually spot-on with his maxims. As it turns out, the maxims subconsciously affect our behavior in any conversation…because we do things to them. When faced with a maxim, we have three choices. We either observe the maxim, flout the maxim, or violate the maxim.
Here’s an example of how we can observe, flout, and violate the maxim of quantity.
Remember, the gist of this maxim is that you try to be as informative as you possibly can, and give as much information as your listener needs, but no more.
Scenario 1: When the maxim of quantity is observed, it sounds like,
Bob: What kinds of cheeses did they have at that new shop around the corner?
Lin: They had blue cheese, swiss, gruyere, I think maybe a couple types of brie? Like normal stuff
In this first scenario, the conversation was quick, and everyone got what they wanted out of it.
“Yes! Got to the end of a conversation without any awkwardness.”
Scenario 2: When the maxim of quantity is violated, it sounds like,
Bob: What kinds of cheeses did they have at that new shop around the corner?
Lin: They had blue cheese, Swiss, gruyere, burrata, American, camembert, brie, soft brie, a harder brie, a cow’s milk goat cheese, a goat cheese gruyere, parmesan, mozzar —
Bob: Ok ok ok, got it 🙂
Lin: *sheepish* Oh, ha, my bad, I’m just overly excited about all the cheeses I guess. We haven’t had such a nice selection since Obama!
In this second scenario, Lin shared a little too much information and violated the maxim…but that actually tells Bob something about how Lin feels about the cheese. Bob had already more or less figured out that Lin was over excited about the cheese selection, because Lin’s violation of the maxim actually communicated extra information to Bob in this case.
“I just love cheese SO MUCH”
Scenario 3: When the maxim of quantity is flouted, it sounds like,
Bob: I gotta hang up in a sec but what kinds of cheeses did they have at that new shop around the corner?
Lin: Nooo can you talk another few minutes?
Bob: Na man I gotta beat this traffic
Lin: 🙁 Wellllll. They had blue cheese, Swiss, gruyere, burrata, American, camembert —
Bob: Ok I get the picture
Lin: 😏 — brie, soft brie, a harder brie, a cow’s milk goat cheese,
Bob: OK GOT IT
Lin: 😏😏 a goat cheese gruyere, Parmasean, mozzar —
Bob: HANGING UP NOW <hangs up>
In this third scenario, Lin actually weaponized the maxim of quantity! By the end of the call, Bob had figured this out, and knew enough not to subject himself to any more of it.
“IF SHE TELLS ME ABOUT ONE MORE VARIETY OF GOAT CHEESE, SH**’S GETTING REAL”
As you can see, nothing about these scenarios is chaotic — it’s actually extremely predictable. In the last two scenarios the cooperative principle was being broken by Lin, but Bob actually figured out that this was happening and used the specifics how how she was violating it to glean extra information about what was happening in the exchange.
Every time someone says something they don’t exactly mean, or every time something seems to goes awry in a conversation (like when someone’s sharing waaay too much info about what they had for dinner last night, or giving you not enough info to answer your question because they assumed you know more about Antarctic penguin migrations than you actually do) it’s because they flouted or violated one of these four maxims. And when they flout or violate a maxim, that behavior tells us some underlying information about the exchange.
As in any kind of communication, context is key, so observing a maxim in one context might actually be violating it in a different context. That’s why it was so important to add in all that extra context in Scenario 3 (the emojis, the scene setting).
Here’s one example of this important context distinction:
If your therapist starts off by asking how you’re doing and you frown and respond with a one-liner, “fine”…that means something important in this context. It probably means you’re not fine. Because you’re at therapy, and the point of therapy is to go a little further than, “I’m fine”. Your therapist would probably follow up by asking lots of questions to get to the heart of why you’re fine or not fine. In this case, you are either violating or flouting the maxim of quantity, and this gives your therapist extra information to decide what to say next to you.
But if the cashier at Albertson’s asks how you’re doing and you frown and say “fine,” it’s unlikely the cashier would start asking lots of questions to try to figure out if you’re actually fine. They’d just know you’re one of the hundreds of “frowning-but-fine” people that they got through the stupid checkout line that day. In this case, even though you responded in exactly the same way that you did with your therapist, you are not violating or flouting the maxim at all — you’re observing it.
“I’d be more fine if I didn’t run the risk of dying from covid every time I get groceries.”
Grice’s maxims as applied to conversation design
So, the question still remains: how can we leverage these maxims and the cooperative principle when designing chatbot or voice experiences?
I do believe the maxims can be useful, but they’re not necessary. And they certainly shouldn’t be treated like an all-purpose best practice checklist of rules to abide by.
As designer Richard Warcheza recently said, referring to the tendency of humans to violate or flout maxims,
“It should seem odd to us to make voice design decisions based on something that was created only to be broken. Right?”
And, as designer Greg Bennet once said,
“Our goal with conversation design is not to be human, but to be natural. Natural enough to facilitate a seamless experience for anyone speaking to the assistant.”
Because all humans are acutely yet subconsciously aware of these unspoken maxims and the various ways we manipulate them, it’s a good idea to assume that your user will be sensitive to when a maxim is violated or flouted. They will assume that any kind of maxim manipulation means something other than what’s being stated in words, just like Bob did in the earlier cheese scenarios.
As a designer, you can actually be sneaky about leveraging this sensitivity to the maxims. James Giangola discussed this in his 2017 Google I/O talk and used the example of a voice experience that helps you track your sleep saying something like,
‘I’ve got a good feeling about your sleep tonight.’
Because people assume the maxims are always in place, they interpret a vague phrase in a way that conforms to their expectations.”
Most importantly, though, a good designer is able to decide whether or not their design observes, flouts, or violates a maxim based on the specific context of the conversational exchange, and whether that observation, flout, or violation is a good thing or a bad thing in the eyes of the user.
So what kinds of context are important to consider?
1. The dialect or the language that people use to speak to your conversational system, and the ways their cultural behaviors impact the way they speak. For example, according to expert designer Madeleine Adkins, a native speaker of Chinese is likely to expect and appreciate a chatbot that can ease into the conversation with multiple turns of small talk, while speakers of American English expect a chatbot to get straight to the point in its very first utterance. Observing the maxims of relevance and quantity in these two contexts will manifest in wildly different ways.
2. The end goal of your user. Are they just trying to transfer money from checking to savings? Or are they trying to get a joke out of your bot? What headspace are they in? A long-winded joke might be observing the maxim of quantity in a lighthearted situation, but completely violating it if the user is trying to file an insurance claim.
3. The personality of your bot. Is your bot warm, friendly, cute, direct, efficient, excited, well-educated, blunt? All of the characteristics you choose for your bot’s personality will impact the definition of a violation, a flout, or an observation of each maxim
4. The power dynamic between the bot and the user. Diana Deibel and Rebecca Evanhoe do an amazing job of explaining this in their book Conversations With Things, (which you can and should now buy!) but essentially they state that not all bot dynamics are created equal. A water company’s chatbot might hold a lot of perceived power over its users if an interaction with it means the difference between the water being turned off or on in that user’s home.
These and many other factors impact the precise definition of observe, violate, or flout in any given conversational context — it’s a moving target. The best conversation designers use this knowledge to their advantage.
When designing a voice or chat experience, being able to predict how your users may perceive certain utterances is a non-negotiable. Here’s the kicker, though.
So many good conversation designers don’t know (or need to know!) that Grice’s maxims are the reason behind their good designs.
In the case of conversation design, understanding the why behind the how can be useful but isn’t necessary. We can easily translate the four aforementioned points into best practices without diving into any theory behind each one:
- Tailor your bot’s language, dialect, conversational style, and choice of vocabulary to the user
2. Make sure that the features of your design accurately deliver on the needs of your users by thinking about what frame of mind a user might be in
3. Develop a strong bot personality that is relatable to your user
4. Think through the perceived power dynamic that may occur between your bot and your user, and adjust your design accordingly
My intention is to illustrate some ways Grice’s maxims can be put to use when designing a conversation — NOT to convince you that to be a _real designer_ you need to become intimately familiar with the maxims so as to prove yourself ✨worthy✨ enough to properly apply the maxims. Because in the same way that lots of graphic designers just have an eye for design, or an instinct that guides them to pick the right colors, even if they’ve never studied Color Theory, I believe that conversation designers don’t really need to know the technicalities behind things like Grice’s maxims.
Humans are just darn good at tuning in to these invisible linguistic forces when we’re communicating with each other. It’s kind of like a secret superpower all of us have, and if your superpowers help you design great conversations without you knowing a single thing about why they’re helping you, then more power to you! As long as you’re willing to think from every angle about how your design could be received, and test with real users to make sure your design delivers something genuinely useful and causes no harm, then you are just as qualified as any linguist to create magical conversations for people.
Huge thanks to Anja de Castro, Jan König, Richard Warcheza, and Margaret Jabczynski for reading through initial drafts of this article. 🙏
Photo #4 from Docorman
Photo #5 from Kutv