3 ms·

Your reasoning is like saying "the canary only disappeared once, and knowing that it can indeed disappear bolsters my confidence in it now that it has returned.

Your reasoning is like saying "the canary only disappeared once, and knowing that it can indeed disappear bolsters my confidence in it now that it has returned." It's plainly wrong. https://en.wikipedia.org/wiki/Exception_that_proves_the_rule https://en.wikipedia.org/wiki/Exception_that_proves_the_rule

kbenson 9y ago
No, it's a matter of reality matching expectations. If you expect that the SLA has very likely been violated at some point, hearing that it has not means you should believe you have just been very likely lied to given your existing knowledge, unless you are considering that statement as a source of truth itself. If you expect that the SLA has very likely been violated at some point, hearing that it has at some point means that the statement confirms to what you already believe to be true given your existing knowledge. That doesn't mean the statement is true, but since it's not obviously conflicting with what you already believe to be true, it at least allows you to believe it is not immediately false. Instead of thinking about it increasing the likelihood of being entirely true, think about it as decreasing the likelihood it's entirely false. Depending on your point of view that may not be much, but it's something.
- jcahill 9y ago
  Absolutely not. >No, it's a matter of reality matching expectations. What an agent tells me is what they choose to tell me. You're describing some sort of luck-based updating via that third party's choice. I don't live in a tinseltown universe full of model trains and animatronic NPCs. None of us do. All reasoning about real-world agents is subject to incomplete information and uncertainty. That much is trivial. More, it's mutually understood to be the case. Also mutually understood: basic world knowledge stemming from the same. These are principles simple enough to be patronizing in written description, yet persistently ignored or misused at implementation-level. Like: 1. Actors are variably susceptible to errors in reasoning under uncertainty 2. Actors are variably skilled at exploiting 1 to modulate 3rd party behavior 3. Actors are variably motivated to make use of 1-2 It follows from the above that allowing an actor to subtly shift your expectations as if you ever held a platonic model of their behavior is simply a cognitive error. There's no way around it. Take the "99.995% SLA" example. In the absence of that figure, would you have assumed a God-Mode level of performance? Clearly not. You can cross all the factors like whether you care about the figure, whether it's above or below average, whether disclosure is standard in this context, … to just enumerate all the cases and see clearly that there's no time when this information is surprising. I mean, just look at a top google hit for SLA⁽¹⁾. You really think a CIO reader is in any way surprised to hear that some metric they negotiated into a contract indeed holds? Or that it doesn't? Continuing: A figure like 99.995% is well within reasonable bounds for any number of business processes, so it's not necessarily false precision here. What it almost definitely is, however, is precision in pursuit of persuasion. There are plenty of industries for which exacting figures at the high end of some performance criterion — manufacturing quality, service availability, measurement accuracy, etc — are essential to informed consumer behavior. Those industries almost universally have norms or regulations setting out certain expectations about what will be found on a specsheet, how units will be tested, how this information will be reported. If not, the spiel is just spiel. Facts and figures as token gestures of fallibility, however, are confidence tricks. I already know you're fallible. You cannot sway this comprehension by reframing around some very likely sort of figure: charm pricing⁽²⁾ and related uses of odd figures are marketing weaselry targeting plebs. To point these things in the direction of clientele is to tell them how much you think of their ability to resist bullshit-fatigue. My feels about whether I'd wanna have a beer with <The Guy>, modulated by his current demeanor toward me or whatever audience he imagines me to be a member of, do not determine his fitness for any high-stakes job. The same is true here. Cloudflare is in the MITM business. Absolutely trusted at no point in time, independent of whatever cost/benefit has gone into the decision to use a MITM. This isn't even defeated by being too big a client to lose: if you were big enough to be a lifeline for cloudflare, you'd have no need for cloudflare. ____________________ ¹ https://www.cio.com/article/2438284/outsourcing/outsourcing-sla-definitions-and-solutions.html https://www.cio.com/article/2438284/outsourcing/outsourcing-... ² http://enwp.org/Psychological_pricing http://enwp.org/Psychological_pricing

kbenson 9y ago

No, it's a matter of reality matching expectations. If you expect that the SLA has very likely been violated at some point, hearing that it has not means you should believe you have just been very likely lied to given your existing knowledge, unless you are considering that statement as a source of truth itself. If you expect that the SLA has very likely been violated at some point, hearing that it has at some point means that the statement confirms to what you already believe to be true given your existing knowledge. That doesn't mean the statement is true, but since it's not obviously conflicting with what you already believe to be true, it at least allows you to believe it is not immediately false. Instead of thinking about it increasing the likelihood of being entirely true, think about it as decreasing the likelihood it's entirely false. Depending on your point of view that may not be much, but it's something.

jcahill 9y ago
Absolutely not. >No, it's a matter of reality matching expectations. What an agent tells me is what they choose to tell me. You're describing some sort of luck-based updating via that third party's choice. I don't live in a tinseltown universe full of model trains and animatronic NPCs. None of us do. All reasoning about real-world agents is subject to incomplete information and uncertainty. That much is trivial. More, it's mutually understood to be the case. Also mutually understood: basic world knowledge stemming from the same. These are principles simple enough to be patronizing in written description, yet persistently ignored or misused at implementation-level. Like: 1. Actors are variably susceptible to errors in reasoning under uncertainty 2. Actors are variably skilled at exploiting 1 to modulate 3rd party behavior 3. Actors are variably motivated to make use of 1-2 It follows from the above that allowing an actor to subtly shift your expectations as if you ever held a platonic model of their behavior is simply a cognitive error. There's no way around it. Take the "99.995% SLA" example. In the absence of that figure, would you have assumed a God-Mode level of performance? Clearly not. You can cross all the factors like whether you care about the figure, whether it's above or below average, whether disclosure is standard in this context, … to just enumerate all the cases and see clearly that there's no time when this information is surprising. I mean, just look at a top google hit for SLA⁽¹⁾. You really think a CIO reader is in any way surprised to hear that some metric they negotiated into a contract indeed holds? Or that it doesn't? Continuing: A figure like 99.995% is well within reasonable bounds for any number of business processes, so it's not necessarily false precision here. What it almost definitely is, however, is precision in pursuit of persuasion. There are plenty of industries for which exacting figures at the high end of some performance criterion — manufacturing quality, service availability, measurement accuracy, etc — are essential to informed consumer behavior. Those industries almost universally have norms or regulations setting out certain expectations about what will be found on a specsheet, how units will be tested, how this information will be reported. If not, the spiel is just spiel. Facts and figures as token gestures of fallibility, however, are confidence tricks. I already know you're fallible. You cannot sway this comprehension by reframing around some very likely sort of figure: charm pricing⁽²⁾ and related uses of odd figures are marketing weaselry targeting plebs. To point these things in the direction of clientele is to tell them how much you think of their ability to resist bullshit-fatigue. My feels about whether I'd wanna have a beer with <The Guy>, modulated by his current demeanor toward me or whatever audience he imagines me to be a member of, do not determine his fitness for any high-stakes job. The same is true here. Cloudflare is in the MITM business. Absolutely trusted at no point in time, independent of whatever cost/benefit has gone into the decision to use a MITM. This isn't even defeated by being too big a client to lose: if you were big enough to be a lifeline for cloudflare, you'd have no need for cloudflare. ____________________ ¹ https://www.cio.com/article/2438284/outsourcing/outsourcing-sla-definitions-and-solutions.html https://www.cio.com/article/2438284/outsourcing/outsourcing-... ² http://enwp.org/Psychological_pricing http://enwp.org/Psychological_pricing