Blogger: Joe Maguire
A few years back, The Humez Brothers and I included this hoary old classic in a book we wrote about language and numbers.
It is September and Fred has just joined the mathematics department as a junior faculty member. To help him feel at home, his office mate, Mary, invites him to accompany the group of math department regulars when they go out for their customary end-of-week beer at the local watering hole. Off they all go, and the conversation is animated and friendly. At one point, Ralph, the abstract algebrist says, apparently out of the blue, “Forty-seven,” at which the regulars laugh uproariously. Fred thinks that this is a bit bizarre but, what with being new to the crowd, says nothing. A bit later, another member of the group, again, apparently out of the blue, says “Twenty-four,” at which, again, everybody laughs wildly. His curiosity getting the better of him, Fred turns to Mary and asks what’s going on. “Oh,” she says, “well, you see, we’ve all been hanging out together for so long that by now we’ve all heard each other’s jokes, so a couple of years ago, we decided to number them all. Now, when somebody wants to tell a joke, he or she simply says the joke’s number and lets it go at that.” Fred ponders this bit of information and, when there is a lull in the conversation says, “Thirty-one.” There is dead silence, and then the conversation picks up again. Fred asks Mary why nobody laughed, to which she replies, “Some people just can’t tell a joke.” (Fittingly, this joke has two variants of which we are aware: When Fred says “Thirty-one,” the assembled company laughs long and loud and with great gusto, various of the group being reduced to tears. When he expresses surprise to Mary that his joke had created such stir, she replies (a) “They’d never heard it told so well” or (b) “They’d never heard that one before.”)
Like all jokes, this one is a narrative. But what’s more, this joke actually comments on narrative—and on the distinction between narrative and structured data.
Jokes are inherently narrative; they are story. The preposterous conceit of this particular joke is that Fred’s colleagues could possibly respond to his utterance (“Thirty-one”) as they would respond to a well-told tale.
Narrative data is characterized by a predictable, sequential access path through the available information (e.g., a novel is narrative because the reader starts at Page 1 and proceeds through the pages in order; a poem is narrative because the stanzas should be read in order). Out of sequence, narrative data can lose significance. A joke won’t work unless the punch line comes last.
By contrast, structured data consists of free-standing, independent facts that retain their value and meaning regardless of the order in which they are encountered. For example, on a Roman soldier’s shopping list, it is immaterial whether the “matching sword-and-shield set” precedes or follows the “new-and-improved, imperial-style centurion’s helmet.”
I sometimes hear it said that structured data is “simpler” than narrative data. Other times, I hear just the opposite. Who’s right?
Everybody’s wrong. That is, anybody making blanket assertions about which kind of data is simpler than the other is missing the point. Structured-data paradigms will appear “too simple” whenever they are used to express narrative-rich data. Likewise, narrative-data paradigms will appear inadequate when they are use to express structured data. It is not a matter of simplicity, but affinity.
Structured data deserves structured representation. If a certain form of data retains its meaning irrespective of surrounding context, that data should be expressed in a structured form. For example, my credit limit is my credit limit, regardless of whether I am buying a steroid-saturated sports car or an anemic motor scooter. Consequently, my credit limit should be expressed in some structured format (e.g., a relational table) so that it can be used in many contexts. It should not been embedded in a document describing my sports-car-seeking mid-life crisis.
Likewise, narrative data deserves narrative representation. Jokes should be told, not merely referred to by their arbitrary numeric identifiers. The joke about Fred and his colleagues illustrates the odd ramifications of applying structured-data paradigms to narrative data.
To be fair, I’ll give an illustration of the opposite mistake—what can happen when structured data is expressed in narrative form.
Unfortunately, this is a true story, so I’ll omit some details and change all names. (And I’m happy to report that I never knew all the details anyway, so don’t ask.) The story: Kathleen, a supervisor at a small business, receives a phone call from Jared, a hiring manager at another business in the same industry. Jared is checking a reference provided by Lee, who once worked for Kathleen. Kathleen speaks extemporaneously about Lee’s good work, which encourages Jared. But in the spirit of thoroughness, Kathleen retrieves Lee’s personnel folder and finds a letter of reprimand within it. Kathleen had forgotten about the letter, because Lee’s transgression occurred several years ago and because Kathleen remembers Lee as a very good employee.
Kathleen reads the letter of reprimand aloud to Jared. The letter spells out the details of Lee’s one-time-only professional misjudgment leading to an unseemly indiscretion. As he hears the story, Jared’s willingness to hire Lee evaporates. Kathleen reads the letter in its entirety, including the last sentence: “Pursuant to policy, this letter of reprimand will be removed from Lee’s personnel file and destroyed after six months.”
What a shame. The more important piece of information is structured data, which could have been more appropriately stored in a named field or column—the name might be something like “Date Of Expiration” or “Removal Date.” However, because the structured data was stored inappropriately as part of a narrative, Kathleen effectively violated her company’s policy and cost Lee a job at another company.
The problem is not that narrative-friendly data paradigms are bad, but that the paradigm was misapplied to non-narrative data.
Burton Group has published research germane to this topic. For more information, see the following documents:
The topic attracts the attention of Burton Group analysts because it is not so simple. For example, some information is neither purely structured nor purely narrative. Our joke about the math department is a case in point: All jokes are inherently narrative, but a joke with three different punch lines has some structured aspects: there is a one-to-many relationship between the setup and the punch lines. Such a structured relationship could accommodate a fourth punch line: After Fred says “Thirty-one,” his new colleagues are repulsed, and the department chair scolds “Hey, that kind of humor is not appropriate for the workplace!”