Assuming that you want to keep the exercise somewhat entertaining, modifying a game like balderdash to start with dilbert-esque executive speak and move towards ever more specific levels could provide a fun, easy to understand method to practice moving up and down in levels of specificity.
As to how this would work in practice, have everyone present get into groups of 5-10 (or however many are sitting at a table) and give everyone a key phrase of execu-speak, such as “Energize the end-user experience.” Then, everyone writes down a short, but more specific description of what the phrase really means. All of the player-generated statements are read aloud to the table (which will also display how vague the initial phrase was, since everyone should have vastly different answers), and the table votes on which is the best written response, and authors get a point for each vote they receive. Then, the exercise is repeated with the phrase the group selected as the best, for several rounds, or until everyone’s responses are approximately the same, showing that a undisputable level of specificity has been reached.
For fun, you could run the exercise in reverse as well, giving a very concrete example and challenging people to go the opposite way, and give a less specific example phrase.
Then, the exercise is repeated [...] until everyone’s responses are approximately the same
In the real world, if the client wants X, it does not matter that ALL the engineers agree that the statement really says Y. Doubly so since engineers and clients often seem to speak two different languages.
It’s far more useful to learn what the CLIENT wants, not what all the engineers THINK she wants. In fact, learning how to come to this sort of false conclusion seems exactly opposite of what Eliezer wants to teach—you now have a group of engineers convinced that they know what this overly-general statement means, despite never having checked back in with the client!
… of course, for a twist, do this, and then tell them that they’re WRONG and the client actually wanted X. Punish them for not thinking to go and actually consult the client. It produces frustration and ill-will (the presenter has just tricked them), but with a receptive audience you do potentially teach a very potent lesson.
What I like about this exercise is, that it doesn’t require any teacher/judge/… who decides what is specific enough (and may thus cause frustration in participants who disagree).
Instead, the exercise ends when everyone agrees on the statements meaning – i.e. when the statement is “specific enough”.
Assuming that you want to keep the exercise somewhat entertaining, modifying a game like balderdash to start with dilbert-esque executive speak and move towards ever more specific levels could provide a fun, easy to understand method to practice moving up and down in levels of specificity.
As to how this would work in practice, have everyone present get into groups of 5-10 (or however many are sitting at a table) and give everyone a key phrase of execu-speak, such as “Energize the end-user experience.” Then, everyone writes down a short, but more specific description of what the phrase really means. All of the player-generated statements are read aloud to the table (which will also display how vague the initial phrase was, since everyone should have vastly different answers), and the table votes on which is the best written response, and authors get a point for each vote they receive. Then, the exercise is repeated with the phrase the group selected as the best, for several rounds, or until everyone’s responses are approximately the same, showing that a undisputable level of specificity has been reached.
For fun, you could run the exercise in reverse as well, giving a very concrete example and challenging people to go the opposite way, and give a less specific example phrase.
In the real world, if the client wants X, it does not matter that ALL the engineers agree that the statement really says Y. Doubly so since engineers and clients often seem to speak two different languages.
It’s far more useful to learn what the CLIENT wants, not what all the engineers THINK she wants. In fact, learning how to come to this sort of false conclusion seems exactly opposite of what Eliezer wants to teach—you now have a group of engineers convinced that they know what this overly-general statement means, despite never having checked back in with the client!
… of course, for a twist, do this, and then tell them that they’re WRONG and the client actually wanted X. Punish them for not thinking to go and actually consult the client. It produces frustration and ill-will (the presenter has just tricked them), but with a receptive audience you do potentially teach a very potent lesson.
What I like about this exercise is, that it doesn’t require any teacher/judge/… who decides what is specific enough (and may thus cause frustration in participants who disagree). Instead, the exercise ends when everyone agrees on the statements meaning – i.e. when the statement is “specific enough”.