Anna glances at a clock that reads 3:00. It is actually 3:00, but the clock stopped exactly 12 hours ago. Does Anna know it is 3:00, according to the safety condition?
AYes — she has a justified true belief, and the safety condition only adds that the method must be reliable in general, which clocks usually are
BYes — she is right, and the safety condition only requires that she couldn't have been wrong given exactly these circumstances
CNo — in nearby possible worlds (she glances slightly earlier or later), her clock-reading method produces a false belief, so her belief is unsafe
DNo — the safety condition requires logical certainty, and it is logically possible the clock stopped recently
The safety condition asks whether Anna could *easily* have been wrong — whether there are nearby possible worlds where her same belief-forming method (reading this clock) yields a false belief. There are: if she glanced a minute earlier or later, the clock would still read 3:00 but the actual time would be different. Her belief is unsafe, so safety correctly withholds the status of knowledge. Option A is the classical JTB analysis that Gettier showed to be insufficient; option D confuses safety with infallibilism.
Question 2 Multiple Choice
Consider a fair lottery with one million tickets. You hold ticket #452,891 and have not yet heard the results. You believe 'my ticket will lose.' Is this belief safe, according to the safety condition?
ANo — if your ticket were the winner, you would still believe it will lose, so you are sensitive but not safe
BYes — in nearly all nearby possible worlds you do lose, so your belief is true in nearly all nearby worlds where you form it
CNo — safety requires that you couldn't be wrong even in slightly different circumstances, and there exists a circumstance (your ticket winning) where you would be wrong
DYes — but only because lotteries are randomly determined, which makes nearby worlds all equally probable
Safety asks: in nearby worlds where you form this belief by the same method, is it true? 'Nearby' means small changes — you still hold ticket #452,891, the draw hasn't happened yet. In the vast majority of nearby worlds, your ticket does lose (probability 999,999/1,000,000). So your belief is safe. Note that safety and sensitivity come apart here: sensitivity asks 'if my ticket were the winner, would I still believe I'll lose?' — and the answer is yes, making the belief *insensitive*. This case illustrates why safety and sensitivity are distinct modal conditions.
Question 3 True / False
The safety condition is sensitive to the specific belief-forming method used, not just to whether the resulting belief happens to be true.
TTrue
FFalse
Answer: True
Safety evaluates whether the method could easily produce a false belief in nearby worlds. Two people can form the same true belief 'It is 3:00' — one by reading a working clock, one by reading a stopped clock — and only the first has a safe belief because the methods differ in their reliability across nearby worlds. A belief can be accidentally true (like reading the stopped clock at exactly the right moment) but still unsafe, because the method would easily mislead in nearby circumstances. This is precisely why safety improves on the simple JTB account.
Question 4 True / False
The safety condition and the sensitivity condition give identical verdicts on whether a belief constitutes knowledge, since both are modal conditions relating belief and truth.
TTrue
FFalse
Answer: False
Safety and sensitivity are both modal conditions but in different directions, and they come apart in important cases. Sensitivity says: if P were false, you wouldn't believe P. Safety says: if you believe P via this method, P is true in nearby worlds. Lottery beliefs are safe (you almost certainly will lose) but insensitive (if you were the winner, you'd still believe you'd lose). Brain-in-vat beliefs are insensitive (if you were a brain in a vat, you'd still believe you have hands) but safe (in nearly all nearby worlds you are not in a vat). Most epistemologists favor safety for tracking knowledge intuitions more reliably.
Question 5 Short Answer
Explain how the safety condition handles a standard Gettier case, and why this represents an improvement over the classical justified true belief analysis.
Think about your answer, then reveal below.
Model answer: In a Gettier case — like reading a stopped clock that coincidentally displays the correct time — the subject has a justified true belief but lacks knowledge because the truth is accidental. The JTB analysis cannot explain this: all three conditions are satisfied. Safety explains it: the belief-forming method (reading a stopped clock) would easily produce a false belief in nearby worlds (one minute earlier or later, the clock still reads 3:00 but the actual time is wrong). The belief is unsafe — the subject 'got lucky' in the actual world, but luck is precisely what safety rules out. A belief counts as knowledge only if the method that produced it is reliably truth-tracking in nearby circumstances, not just accidentally true in this one.
The safety condition essentially says: you know P only if your being right about P is not a matter of luck in the relevant sense. Gettier cases are paradigm cases of epistemic luck — you are right, but you could very easily have been wrong using the very same method. Safety captures this by looking at nearby possible worlds and asking whether the method is reliable there. The JTB analysis has no modal component and so cannot distinguish knowledge from lucky true belief.