The Structure of Open Secrets
given on 1 December 2023
at Brandeis
Sam Berstler
MIT Philosophy
berstler@mit.edu
This talk will involve brief discussions of:
substance abuse
sexual misconduct
spoilers for House of the Dragon
Wilcock: Well, okay, umm…And this girl, has she done something to make you, ummmm, do you suspect that she’s done something to harm the company?
Wilcock: I don’t think you have anything to worry about.
Jane: Hmm?
Wilcock: You’re not his type.
In The Assistant, they do talk about the boss’ conduct, and they don’t literally ignore. What they don’t do is openly acknowledge it.
They do say…
They don’t say…
3. Your boss is a sexual harasser, but we don’t talk about that around here. We all just kind of tolerate it.
4. If you file this complaint, we’ll make sure you never work in this industry again.
Anonymous whistleblower: “It was amazing. Here we were dumping poison into the environment, and nobody wanted to talk about it, as if talking about it would make it real. Well, it was real all right, but we went around pretending it wasn’t. I thought I was going crazy, like it wasn’t happening. You think I’m some kind of hero ‘cause I blew the whistle. The only reason I spoke up is because I didn’t want to go crazy.”
Observation: In conversations, we often share too much knowledge. In these cases, we rely on non-acknowledgment practices in order to coordinate our interaction.
This information might be
…embarrassing
…psychologically distressing
…a source of conflict
…legally regulated
…otherwise private
16
The essential function of the boundary between what is acknowledged and what is not is to admit or decline to admit potentially significant material into the category of what must be taken into consideration and responded to collectively by all parties in the joint enterprise of discourse, action, and justification that proceeds between individuals whenever they come into contact. If something is not acknowledged, then even if it is universally known, it can be left out of the consideration of the collective social process…
Nagel (1998: 12)
Open secrecy is a special kind of non-acknowledgement practice. It is characterized by its iterative structure.
If p is an open secret, then there’s a social norm not to acknowledge that p, and this norm is itself an open secret.
The big question
Why do some groups seem to “turn in a blind eye” or “remain in denial” about wrongdoing within the group?
Some familiar explanations
1. Moral failing
2. Individual epistemic dysfunction
3. Collective epistemic dysfunction
The big question
Why do some groups seem to “turn in a blind eye” or “remain in denial” about wrongdoing within the group?
Some familiar explanations
1. Moral failing
2. Individual epistemic dysfunction
The big question
Why do some groups seem to “turn in a blind eye” or “remain in denial” about wrongdoing within the group?
How do they work and what insights do they provide about the nature of discourse structure?
Some familiar explanations
1. Moral failing
2. Individual epistemic dysfunction
Some examples
Torture
At war with sectarian rebels, a government initiates a program to “intensively interrogate” detainees. If asked whether they are torturing detainees, the government, interrogators, and guards will deny that they are. When talking to each other, they also all describe what they are doing as intensive interrogation. But everyone involved in the program understands that what they are really doing is torture.
(adapted from Cohen 2001: 107)
Alcoholism
A mother, Eloise, arrives home with her son, Corey. Eloise’s husband, Jerry, is asleep on the couch. The house is a mess. Alcohol bottles are strewn everywhere. What’s happened is obvious to both Eloise and Corey: Jerry drank too much and passed out. Eloise and Corey silently clean up the house. The next day, they don’t talk about what happened. This scene repeats over and over again throughout Corey’s childhood.
(adapted from Zerubavel 2006: 7)
Unfair will
Amelia’s grandmother Rose recently passed away. Rose left a will that unfairly advantages her youngest and favorite son, Chet. No one, least of all Chet, thinks that Chet deserves the extra money. He wasn’t a good son to Rose, he has much more money than the rest of the family, he didn’t invest in Rose’s end of life care, and generally he’s a jerk. Still, the rest of the family knows there’s no point in confronting Chet about it, because it will only fruitless arguments.
After the funeral, Chet says, “I’m glad to see that Rose left such a fair will. It would be a shame to squabble about money.” Without any sarcasm in her voice, Amelia replies, “That’s right, Uncle Chet, thank goodness the will was fair.” The rest of the family keeps saying this sort of thing for years.
(paraphrased from Michaelson 2018: 184)
Illegitimate heir
In the fictional country of Westeros, the children of the crown princess, Rhaenyra, are obviously illegitimate. In this particular fantasy world, two parents with silvery blond hair (associated with a particular magical dynasty) always have children with silvery blond hair. Rhaenyra and her husband both have this hair. Their children have brown hair. Moreover, while Rhaenyra is racially white, her husband is racially Black. None of Rhaenyra’s children have racially Black features. Despite some quiet grousing and bawdy jokes, there is a long period of time during which the nobles almost universally pretend that there is no reason to question the legitimacy of Rhaenyra’s children.
(House of the Dragon 2022)
Meta-linguistic open secrecy
p: “Torture” is a way to refer to what we are doing.
p is an open secret.
Evidential open secrecy
p: There’s good evidence that Rhaenyra’s children are illegitimate.
Merely moral open secrecy
p. It’s wrong to sexually harass your employees.
(Where, Smith sexually harasses his employees is not also an open secret.)
Simple open secrecy
p: Harvey sexually harasses his employees.
Some features of
open secrecy
Open secrecy feature 1: When p is an open secret in a group g, there is a social norm in g not to acknowledge that p.
Open secrecy feature 1: When p is an open secret in a group g, there is a social norm in g not to acknowledge that p.
The hallmark of social norms is the group’s pattern of enforcing the norm, through sanctioning norm-breakers and rewarding norm-conformers.
Open secrecy feature 2: When p is an open secret in a group g, there is a social norm in g not to acknowledge that p when interacting with members and non-members of g.
Open secrecy feature 2: When p is an open secret in a group g, there is a social norm in g not to acknowledge that p when interacting with members and non-members of g.
An omerta norm is not an open secrecy norm.
Omerta norms
A group g has an omerta on p iff there is social norm in g, such that members of g may not acknowledge that p to non-members of g.
Open secrecy feature 3: When p is an open secret in a group g, then p is close to common knowledge in g.
Open secrecy feature 3: When p is an open secret in a group g, then p is close to common knowledge in g.
Common Knowledge
A and B jointly know that p iff
A and B know that p,
A and B know that A and B know that p,
A and B know that A and B know that A and B know that p,
et cetera..
(Lewis 1969)
Common Knowledge
A and B jointly know that p iff
A and B know that p,
A and B know that A and B know that p,
A and B know that A and B know that A and B know that p,
et cetera..
(Lewis 1969)
Why knowledge and not belief?
I’m assuming the knowledge norm on action (Hawthorne and Stanley 2008). The basic idea: open secrecy solves a practical problem. If our shared belief that p doesn’t give us reason to act on p, then we don’t face the relevant practical problem.
Open secrecy feature 3: When p is an open secret in a group g, then p is close to common knowledge in g.
Common Knowledge
A and B jointly know that p iff
A and B know that p,
A and B know that A and B know that p,
A and B know that A and B know that A and B know that p,
et cetera..
(Lewis 1969)
Why knowledge and not belief?
I’m assuming the knowledge norm on action (Hawthorne and Stanley 2008). The basic idea: open secrecy solves a practical problem. If our shared belief that p doesn’t give us reason to act on p, then we don’t face the relevant practical problem.
Open secrecy feature 3: When p is an open secret in a group g, then p is close to common knowledge in g.
A group treats p as an open secret iff not p, the group has something close to joint belief that p, and the group has all the other relevant open secrecy norms.
Bookkeeping on feature 3
Open secrecy feature 4: When p is an open secret in a group g, then p.
Open secrecy feature 4: When p is an open secret in a group g, then p.
Why?
1. If p is an open secret in g, then g has something
like joint knowledge that p.
2. If g jointly knows that p, then p.
3. So if p is an open secret in g, then p.
Open secrecy feature 4: When p is an open secret in a group g, then p.
Why?
1. If p is an open secret in g, then g has something
like joint knowledge that p.
2. If g jointly knows that p, then p.
3. So if p is an open secret in g, then p.
This accords with natural language data.
You can derive an iteration principle just from the feature 4 (factivity) and some plausible assumptions, but it’s strictly weaker than the iteration principle that interests me.
You can derive an iteration principle just from the feature 4 (factivity) and some idealizing assumptions, but it’s strictly weaker than the iteration principle that interests me.
Assumption 1
If p entails q, then if I acknowledge that p, I also acknowledge that q.
Assumption 2
Assumption 1 is common knowledge.
Assumption 3
If p entails q, and a group has something like common knowledge that p, they have something like common knowledge that q.
You can derive an iteration principle just from the feature 4 (factivity) and some idealizing assumptions, but it’s strictly weaker than the iteration principle that interests me.
Assumption 1
If p entails q, then if I acknowledge that p, I also acknowledge that q.
Assumption 2
Assumption 1 is common knowledge.
Assumption 3
If p entails q, and a group has something like common knowledge that p, they have something like common knowledge that q.
Then, if p is an open secret in g, it’s an open secret in g that p is an open secret in g.
Open secrecy feature 5: When p is an open secret in a group g, then there is a social norm in g not to acknowledge that p, and this norm is itself an open secret.
Open secrecy feature 5: When p is an open secret in a group g, then there is a social norm in g not to acknowledge that p, and this norm is itself an open secret.
Notes
Open secrecy feature 5: When p is an open secret in a group g, then there is a social norm in g not to acknowledge that p, and this norm is itself an open secret.
Notes
Unlike the weaker iteration principle, this iterative principle prohibits acknowledging the norm even if you do not acknowledge that the norm targets true information:
We do not talk about the allegations against Harvey around here.
Amelia conforms to the norms here:
Danny: Can we just talk openly about whether or not the will is fair? I’m tired of all this pretending.
Amelia: No one is pretending. We’ve been talking openly.
Don’t say…
1. Corey, we don’t discuss your dad’s alcoholism.
2. I’m firing you, because you violated our rule against discussing the allegations against Harvey.
Many non-acknowledgment practices lack this iterative structure.
Many non-acknowledgment practices lack this iterative structure.
Closing the book
A group g has a closing-the-book norm on p just in case there is a social norm or agreement in g, such that members of g may not discuss whether p.
Agree to disagree
A family is fighting so much about whether the president’s latest immigration policy is ethical that they all decide not to talk about it. When someone forgets and brings it up, her family members might say:
1. We agreed not to talk
about this!
Many non-acknowledgment practices lack this iterative structure.
Faculty meeting
In a hiring meeting, the faculty agree to exclude considerations about where a candidate earned her PhD. A faculty member says:
1. Can we revisit our agreement not to discuss this? I think the fact that this candidate earned a PhD from MIT is important. (It shows that they’re probably a possible worlds fanatic!)
Exclusion norm
A group g has an exclusion norm on p just in case there is a social norm or agreement in g, such that: for some contexts c, even if members of g know that p, they may not treat p as a reason for action.
Many non-acknowledgment practices lack this iterative structure.
Vulgarity (not open secret euphemism)
A: In this example, Rhaenyra fucked
someone who wasn’t her
husband.
B: Yes but this is a professional
context. You should really say
“had sex with.”
Polite euphemisms
A group g has a polite euphemism norm on some G just in case in certain contexts in g, members may not use some set of strings S in order to refer to G.
Torture (open secret euphemism)
A: We’re torturing some folks, aren’t
we?
B: Yes but you should really say
“intensively interrogate.”
First pass analysis
P is an open secret in a group g iff
(a) p is close to common knowledge in g,
(b) there is a norm in g not to acknowledge that p,
(c) (b) is an open secret.
P is an open secret in a group g iff
(a) p is close to common knowledge in g,
(b) there is a norm in g not to acknowledge that p,
(c) (b) is an open secret.
Hypothesis 1: you acknowledge that p iff you assert that p.
Problem: this is too weak! It would allow members to say…
A comedic example of someone who asserts not p but simultaneously presupposes p:
Larry: See, look, if it were defamation, there would have to be
someone I was defaming. All right, all right, let’s keep it
simple. I could pretend the money never appeared. That’s
not defaming anyone.
Clive’s father: Yes. And passing grade.
Larry: Passing grade?
Clive’s father: Yes.
Larry: Or you’ll sue me--?
Clive’s father: For taking money.
Larry: So he did leave the money?
Clive’s father: This is defamation.
Larry: It doesn’t make sense. Either he left the money or he
didn’t…
Clive’s father: Accept a mystery.
Second pass analysis
P is an open secret in a group g iff
(a) p is close to common knowledge in g,
(b) there is a norm in g not to acknowledge that p,
(c) (b) is an open secret.
Hypothesis 2: you acknowledge that p iff you presuppose that p.
Some tools
A accepts that p for the sake of a conversation c iff she acts as if p for the sake of c.
p is common ground in c just in case A and B jointly believe that they accept that p for the sake of c.
A presupposes that p for the sake of a conversation c iff they accept that p is common ground for the sake of c.
(Stalnaker 2002, cf. 2014)
A accepts that p for the sake of a conversation c iff she acts as if p for the sake of c.
p is common ground in c just in case A and B jointly believe that they accept that p for the sake of c.
A presupposes that p for the sake of a conversation c iff A accepts that p is common ground for the sake of c.
(Stalnaker 2002, cf. 2014)
The “playing ground” of the conversation; target of our speech acts; what we rely on in order to coordinate.
A accepts that p for the sake of a conversation c iff she acts as if p for the sake of c.
p is common ground in c just in case A and B jointly believe that they accept that p for the sake of c.
A presupposes that p for the sake of a conversation c iff A accepts that p is common ground for the sake of c.
(Stalnaker 2002, cf. 2014)
A is “taking it for granted” or “treating it as background.”
A accepts that p for the sake of a conversation c iff she acts as if p for the sake of c.
p is common ground in c just in case A and B jointly believe that they accept that p for the sake of c.
A presupposes that p for the sake of a conversation c iff A accepts that p is common ground for the sake of c.
(Stalnaker 2002, cf. 2014)
A can accept p when she doesn’t believe p.
A can fail to accept p when she believes p.
Joint beliefs
Common ground
Common Ground
(for c)
Joint
beliefs
Common Ground
(for c)
Common Ground
(for c)
Joint
beliefs
Common Ground
(for c)
Joint
beliefs
Pretense
Simple conversation
Assuming
Ignoring
Some speakers, S…S*, are not acknowledging p in a conversation c iff
(a) S…S* jointly know in c that p
(b) S…S* are not presupposing that p in c.
Non-Acknowledgment Attempt 2
Some speakers, S…S*, are not acknowledging p in a conversation c iff
(a) S…S* jointly know in c that p
(b) S…S* are not presupposing that p in c.
Non-Acknowledgment Attempt 2
This works well with some cases but requires us to give up a central assumption about the common ground. It doesn’t work well with other cases.
Aemond: To the health of my nephews. Jace. Luke. And Joffrey. Each of
them handsome, wise….strong…
[Implicature: Their father is Harwin Strong [not their legal father]]
Alicent: Aemond!
Aemond: Come! Let us drain our cups to these three strong boys.
[[Implicature: Their father is Harwin Strong [not their legal
father]]
Jace: I dare you to say that again.
Aemond: Why? ‘Twas only a compliment. Do you not think yourself
strong?
(Jace and Luke attack Aemond and his brothers)
Alicent: Why would you say such a thing in front of all these people?
Aemond: I was merely expressing how proud I am of our family, Mother.
Though it seems my nephews aren’t quite as proud of theirs.
Aemond: To the health of my nephews. Jace. Luke. And Joffrey. Each of
them handsome, wise….strong…
[Implicature: Their father is Harwin Strong [not their legal father]]
Aemond is exploiting the fact that his insinuation is only recoverable if you presuppose the open secret:
Open secret: Jace, Luke, and Joffrey look overwhelmingly like the
children of Harwin Strong.
He has implausible deniability for insinuating this. No one can openly sanction him for his insinuation without herself violating the open secrecy norm.
p is common ground in c just in case A and B jointly believe that they accept that p for the sake of c.
A presupposes that p for the sake of a conversation c iff A accepts that p is common ground for the sake of c.
(Stalnaker 2002, cf. 2014)
The “playing ground” of the conversation; target of our speech acts; what we rely on in order to coordinate.
Problem!
We need to give up this assumption:
If A and B can rationally rely on joint knowledge in p that c, then
p is common ground in c.
Informally: The common knowledge doesn’t entail all the information that’s in principle available to coordinate communication.
🡪 Some philosophers suggested that this is the right way to go (Stalnaker 2014, Camp 2018).
Deeper problem!
This only works well for what I’ll call perfect open secrets.
Deeper problem!
This only works well for what I’ll call perfect open secrets.
It also doesn’t explain why Alicent’s rebuke is better than her hypothetical rebuke in (2), since they both violate the open secrecy norm.
father is Harwin Strong in front of all of
these people.
Perfect open secrecy: aims to create something that approximates ignorance of p.
Imperfect open secrecy: doesn’t aim to create something that approximates ignorance of p.
Perfect open secrecy: aims for groups not to jointly access the information that p.
Imperfect open secrecy: emerges when groups have a continual need to jointly access the information that p.
The unfair will
The illegitimacy of Rhaenyra’s children
Jerry alcoholism: need to work together to hide the bottles
Weinstein: need to work together to intimidate journalists
Dumping poison into the environment: need to work together to dump it
Torture (“what we are doing is called ‘torture’”): need to torture
P is an open secret in a group g iff
(a) p is close to common knowledge in g,
(b) there is a norm in g not to acknowledge that p,
(c) (b) is an open secret.
Hypothesis 3: when p is an imperfect open secret, you must pretend (presuppose) that not p.
Good news
Act as if Weinstein isn’t a predator.
Act as if Weinstein isn’t.
Notice that this is what actors on stage do. They presuppose (act as if) they’re Hamlet and Horatio and presuppose (act as if) they’re actors playing Hamlet and Horatio.
How do we know?
For example, they work together to turn out
towards the audience.
Good news
In cases in which it’s edge case whether the speaker has acknowledged the open secret, it’s because it’s an edge case whether the speaker is maintaining a pretense that not p.
We can also say that when a speaker cannot maintain a perfect open secret, the norm instructs her to maintain an imperfect open secret.
Good news
The pretense is mandatory, which means that every move in the conversation must update the pretense. This explains why (3) is unacceptable.
First assertion updates the pretense.
Second assertion updates the non-pretense without updating the pretense.
Bad news
In normal conversational pretenses, we can openly mark our pretense as a pretense, through special intonations, facial expressions, or obviously exaggerated behavior.
This is not permissible in open secrecy, and we don’t have clear resources to say why.
81
We need more structure in our theory. Here, I’m inclined to explore Erving Goffman’s sociological notion of the definition of the social situation.
82
Definition of the social situation: we can define a social situation as real even if we rely on a pretense in order to do so. This is what we do in open secrecy cases but not in jokey pretenses.
83
We use a “secondary track” (here, the second common ground) in order to preserve the seeming-reality of the social situation. What constitutes the secondary track is often negotiated. But what happens in the secondary track doesn’t directly damage our social standing (face) or undermine the integrity of the primary social interaction.
P is an open secret in a group g iff
(a) p is close to common knowledge in g,
(b) there is a norm in g not to acknowledge that p,
(c) (b) is an open secret.
Hypothesis 3: when p is an imperfect open secret, you must define the situation as one in which not p.
Big question: Why do we comply with open secrecy norms, even when it should be obvious that they’re morally perverse?
3. Sometimes, we think that that perpetrator
did do something wrong but that the open
secret is justifiably protecting her privacy. So we think that we ought to comply with the open secret.
Big question: Why do we comply with open secrecy norms, even when it should be obvious that they’re morally perverse?
FUN COMBO!
4. We’re “really” conforming out of fear
of retaliation. But we tell ourselves that we
believe that open secrecy is norm is justified,
since that’s more psychologically bearable.
why think we’re conforming due beliefs about privacy?
What people say
“Don’t air our dirty laundry.”
“You need to be discrete.”
“I don’t do gossip.”
The “Going to Vegas” phenomenon
People are most willing to talk about open secrets in contexts where privacy norms are globally relaxed or regularly flouted. These include:
risqué or edgy joking
drunken conversations
highly intimate gossip sessions
Big question: Why do we comply with open secrecy norms, even when it should be obvious that they’re morally perverse?
1. Because they’re so hard to learn, knowledge of the open secret and its attendant norms becomes an effective way to demarcate insiders from outsiders.
2. Some people may want to keep the open secret going to preserve their status as an insider in a group.
Why should I care?
Tell me why I’m wrong.
I am slow on email, but emails are always welcome:
berstler@mit.edu