One advantage of this over map-territory correspondence is that it explains the asymmetry between map and territory. Mutual information is symmetric. So why is the map about the territory, but not the other way around? Because the map has been optimized to fit the territory, not the other way around. (“Fit” in the sense of carrying high mutual information, which can be decoded via some specific intended correspondence—a symbolic language.)
loose thought: I wonder what would be the active inference crowd’s take on this
In this example, teleosemantics vs probabilistic implication matches fairly well with literal meaning vs connotation. However, there are some cases which present more difficulties:
A more socially savvy communicator will understand the connotations of their speech, and optimize for these as well.
One of my main criticisms of the probabilistic-implication account of meaning was its inability to properly define lying. However, this also appears to be a problem for the current account!
My thoughts on this which (I think) (at least partially) align with yours but also take a different perspective
Re (2) lying:
Say sender S is sending a message M to receiver R and M is a lie in the sense that S sends M to R because S believes that it will sway R’s beliefs away from (what S thinks sees as) truth. Here, M denotes something else to S and R. To S, M denotes that R will start believing some untrue proposition P (or allocate more probability mass on P) upon receiving M. S has some M-related metadata/background assumptions that prevent them from taking M at face value. R does not have those assumptions, so they do take M at face value.
S optimized M to mean (to S themself) “R will start believing proposition P more upon receiving M”. This also means that (if S’s assumptions that caused them to send X to R are correct) it will mean something else to R (namely “P”).
Re (1) connotation optimization:
I think there are (at least) two ways we can disentangle this.
First, if the sender optimizes the connotations but the receiver assumes that only the denotations were optimized, then we have again the distinction between what the signal means to the sender and to the receiver.
Second, if we have communication where both sides assume there may be some non-trivial optimization of connotations for (honest) communicative purposes, we have the distinction between the communication channels that are typically optimized to convey the information (as assumed by the society/linguistic community) and the communication channels that are actually optimized to convey the message in the case of these two speakers.
(This brings to my mind the distinction between genetic inheritance and other inheritance systems. In particular, both words/literal meaning/denotation and genes seem to be more unambiguously decodable than the other stuff. If the literal/denotationary communication channel is symbolic (as is the case with natural language), then we also have the similarity in that both genetic inheritance and language are ~discrete, which not only makes them easier to analyze but also makes them naturally preferable channels for communication (inheritance is kinda communication between generations).)
loose thought: I wonder what would be the active inference crowd’s take on this
My thoughts on this which (I think) (at least partially) align with yours but also take a different perspective
Re (2) lying:
Say sender S is sending a message M to receiver R and M is a lie in the sense that S sends M to R because S believes that it will sway R’s beliefs away from (what S thinks sees as) truth. Here, M denotes something else to S and R. To S, M denotes that R will start believing some untrue proposition P (or allocate more probability mass on P) upon receiving M. S has some M-related metadata/background assumptions that prevent them from taking M at face value. R does not have those assumptions, so they do take M at face value.
S optimized M to mean (to S themself) “R will start believing proposition P more upon receiving M”. This also means that (if S’s assumptions that caused them to send X to R are correct) it will mean something else to R (namely “P”).
Re (1) connotation optimization:
I think there are (at least) two ways we can disentangle this.
First, if the sender optimizes the connotations but the receiver assumes that only the denotations were optimized, then we have again the distinction between what the signal means to the sender and to the receiver.
Second, if we have communication where both sides assume there may be some non-trivial optimization of connotations for (honest) communicative purposes, we have the distinction between the communication channels that are typically optimized to convey the information (as assumed by the society/linguistic community) and the communication channels that are actually optimized to convey the message in the case of these two speakers.
(This brings to my mind the distinction between genetic inheritance and other inheritance systems. In particular, both words/literal meaning/denotation and genes seem to be more unambiguously decodable than the other stuff. If the literal/denotationary communication channel is symbolic (as is the case with natural language), then we also have the similarity in that both genetic inheritance and language are ~discrete, which not only makes them easier to analyze but also makes them naturally preferable channels for communication (inheritance is kinda communication between generations).)