How is that not analogous to a wall safe with my journal in it?
To simplify it, let's imagine a one time pad encryption. Does the information exist, or does it only exist in potentia? I think there is a good argument in favor of defining it only in terms of potential information as any one time pad can represent any information of the same length (or smaller if we accept padding).
With Schrödinger's cat experiment we could create a similar setup for the journal but it's questionable if we can still apply the perspective that the journal only exist in potentia. From the cat's perspective it knows if it exists or not. The encrypted information however has a more metaphysical environment and it is more questionable if it can be said to have a similar perspective.
The way programs detect if the encryption is successful is usually by looking at the first bits of information with the assumption that random collisions are unlikely to produce an expected pattern. Not all decryption systems does this however and some just give you the data as produced by the given key.
Both are however just technical details in how to turn the potentia of the random-like encryption data into information.
If it's just random bits that you need a one-time pad to decode, then there isn't any information without the decryption key.
If you're encrypting a hard drive, most encryption methods give you full certainty that you've correctly decrypted the text, in the same way that you'd have full certainty that you've correctly opened a safe and found the journal inside.
I was thinking about mention it before when I wrote the above comment but it was already becoming a lengthy comment.
Truecrypt (now Veracrypt) is one of the more popular disk encryption software and was part of at least one US lawsuit in regard to revealing passwords. Truecrypt support a technique called hidden drives. The technique use the fact that free space is indistinguishable from encrypted data, so an attacker can never be fully certain if they have decrypted the whole data or just part of it.
A older and similar concept was/is utilized by Freenet project. Here the data get one-time pad encrypted using existing encrypted data blocks of same size. Each encrypted block then becomes both the key and data from the perspective of the encryption scheme, and the same block can be reused multiple times as one side of the operation for any given number of decrypted data. In order to decrypt a given file you need to first download the map that identify which blocks represent both sides of the one-time pad encryption, then the blocks which combined are twice the size of the decrypted data, and then do the operation. Freenet theorized that since any block could be the key/data for any other block you could never be certain of what information you have stored by looking at a single block. The block is just information in potentia.
It is also important to understand that its not the journal itself that get subpoena. In the later case it is the wall case, with the government arguing that all locked wall cases must contain an unlocked wall case. The conclusion of the existence of an unlocked wall case is thus a forgone conclusion, with the content within being irrelevant to the argument.
The court in this case looked at this and said "The Commonwealth is seeking the password, not as an end, but as a pathway to the files being withheld". This basically mean that they don't accept the argument that the government can request the password based on the simple fact that an encrypted disk exist. Thus the focus is changed away from the container and onto the information within, and here the court do not think the government has enough information to prove a forgone conclusion.
It is further impossible to distinguish between won't and can't unlock.
Even if believed that you can subpoena the content of your skull you arrive at a situation where every criminal "can't remember"
The article didn't really argue something like that. It only reported on a court case and really didn't say if they agree with the decision or not.