In today’s short post, I shall argue that (a) there are at least some messages which we can identify as the product of an intelligent agent, regardless of their linguistic and social context, and (b) there is no context in which it would be reasonable for us to conclude that a message visible to everyone was a hallucination.
What prompted this discussion
In a post titled Signature in the cell?, Professor Edward Feser argued that no message, in and of itself, could warrant the inference that it was the product of an intelligent agent, without a knowledge of the context of the message. Referring to the hypothetical scenario in which a “Made by Yahweh” message was discovered in every human being’s cells, Feser wrote:
If we’re to judge that Yahweh, rather than extraterrestrial pranksters, hallucination, or some other cause, was behind such an event, it is considerations other than the event itself that will justify us in doing so.
The reference to “hallucination, or some other cause” (presumably a natural one) as a possible explanation for the “Made by Yahweh” message in every human being’s cells led me to infer that Feser was acknowledging the legitimacy of a hyper-skeptical stance here – a position for which I criticized him in a subsequent post. Feser wrote a follow-up post in reply, in which he clarified his position:
I neither said nor implied that it would be “perfectly rational” to interpret phrases like the ones in question [e.g. the “Made by Yahweh” message in every cell – VJT] as hallucinations or as something other than a product of intelligence… What I said is that determining what to make of such weird events would crucially depend on epistemic background context, and that if we concluded that God was responsible (as of course we well might), then that epistemic background context would be doing more work in justifying that judgment than the weird events themselves would be.
In a comment attached to a recent post on Professor Feser’s Website, I pressed him to answer two simple questions of mine:
…[A]s an ID theorist, I happen to think it’s absolutely obvious that we can identify some messages as the work of an intelligent designer, regardless of context… From my reading of your [earlier] post, it seemed to me that you were saying that context was essential when drawing the inference that a message was the work of an intelligent agent. I would profoundly disagree.
I’d like to bury the hatchet, so I’ll ask you two questions:
1. Do you agree that if a message saying “Made by _____” were discovered in every human’s cells, it would be irrational to explain away the discovery as a mass hallucination, regardless of whether the message referred to God, Quetzalcoatl, or Steve Jobs as its author?
2. Do you agree that if the message were suitably long and specific (say, 100 characters of perfectly grammatical English with no repetition), it would be irrational not to ascribe the message to an intelligent agent, regardless of the message’s context?
As we’ll see below, Feser’s answer to both questions was “No.”
Feser replied:
…[O]ther readers have already pointed out what is wrong with your questions. Of course context would be relevant to interpreting such messages. Now, I can easily imagine contexts in which it would be extremely unreasonable to say “Oh, this is a hallucination” and I can easily imagine contexts in which it would not be. If we describe various possible contexts in enough detail, we can certainly see how they would make a clear answer possible. That’s why there’s nothing remotely skeptical about what I said. Give us a specific context and sure, we can decide “This suggested interpretation is just indefensible” or “That suggested interpretation is extremely plausible.” But it’s silly to say “Let’s abstract from all context and then ask what the most probable source of the phrase is.” As Mike Flynn pointed out above, there’s no such thing as the most probable source absent all context.
Feser continued:
BTW, Vincent’s attempt to wriggle out of the problem context poses for his position is like certain point-missing attempts to solve the “commonsense knowledge problem” in AI [artificial intelligence – VJT]. As Hubert Dreyfus argues, it makes no sense to think that intelligence can be reduced to a set of explicitly formulated rules and representations, because there are always various context-dependent ways to interpret the rules and representations. To say “Oh, we’ll just put the ‘right’ interpretation into the rules and representations” completely misses the point, since it just adds further rules and representations that are themselves subject to alternative context-dependent interpretations.
Vincent is doing something similar when he tries to come up with these goofy examples of really long messages written in the cell. It completely misses the point, because that’s just further stuff the import of which depends on a larger context. It also completely misses the point to shout “Skepticism!”, just as an AI defender would be completely missing the point if he accused Dreyfus of being a skeptic. There’s nothing skeptical about it. We can know what the context is and thus we can know what the right interpretation is; we just can’t know the right interpretation apart from all context.
What is a context, anyway?
Remarkably, nowhere in his post does Professor Feser attempt to define what he means by a context – a curious omission. So I’m going to go with a standard dictionary definition: “the circumstances that form the setting for an event, statement, or idea, and in terms of which it can be fully understood and assessed.” I should mention that there is another definition for context: “the parts of something written or spoken that immediately precede and follow a word or passage and clarify its meaning.” However, in the case under consideration, we are looking at a short isolated message, with nothing preceding or following it. So the questions we need to confront are: do we need to attend to “the circumstances that form the setting” for the purported message, in order to rationally conclude that it is (a) not a collective hallucination we are all having, and (b) from an intelligent source? Feser contends that we do, and I maintain that we do not.
Feser’s absurd epistemic claim: there are some contexts in which hallucination may be a reasonable explanation for the discovery of a purported message in every human’s cells
I’d like to go back to a remark Feser made above:
Of course context would be relevant to interpreting such messages. Now, I can easily imagine contexts in which it would be extremely unreasonable to say “Oh, this is a hallucination” and I can easily imagine contexts in which it would not be.
What Feser is saying here is that there are at least some contexts in which it would not be unreasonable [i.e. it might be reasonable] for us to conclude that a purported message discovered by scientists in every human being’s cells was in fact a hallucination. This, I have to say, is outright nonsense.
In order to see why it’s nonsense, let’s imagine a scenario which is as generous to Professor Feser’s case as it is possible to be. Let’s suppose that a worldwide magnetic storm is playing havoc with people’s brains, causing them to hallucinate. It has been claimed that magnetic stimulation of the brain can trigger religious hallucinations, although the evidence for this claim is very thin. But let’s suppose for argument’s sake that this claim is true. During the magnetic storm, some scientists suddenly announce the discovery of a “Made by Yahweh” message in every human being’s cells. Other scientists around the world rush to confirm the claim. Could they all be seeing things in their laboratories? Could mass hallucination be a rational explanation for this sudden discovery of what appears to be a message in our cells?
No, it couldn’t – unless all the world’s scientists have not only started hallucinating, but lost their ability to reason, as well. But that wasn’t the scenario envisaged by Feser: his assertion that he can imagine at least some contexts where it would not be unreasonable to conclude that a purported message was a hallucination presupposes that the people drawing this conclusion still possess the use of reason, even in these far-fetched contexts.
One obvious way in which scientists could confirm that the message was real – even during a magnetic storm that was playing havoc with their perceptions – would be to use double-blind testing, with a control sample of similar-looking cells (say, synthetic cells, or perhaps cells from another species) that did not contain the “Made by Yahweh” message. (A control sample of synthetic cells might contain no message at all, or alternatively, a different message – “Made by Craig Venter” – might be inserted into the cells.) If testing on different scientists produced consistent results – e.g. if they all reported seeing the same message in the same cells – then the hallucination hypothesis would be decisively ruled out, as an explanation.
Interpretation is not the same thing as decoding: why the commonsense knowledge problem is irrelevant to the Intelligent Design project
In his reply to my questions, Feser alluded to the work of AI researcher Hubert Dreyfus, who in a book titled Mind over Machine (Free Press, 1986) which he co-authored with Stuart Dreyfus, defined the commonsense knowledge problem as “how to store and access all the facts human beings seem to know” (1986, p. 78). As Wikipedia notes, “The problem is considered to be among the hardest in all of AI research because the breadth and detail of commonsense knowledge is enormous.”
As we’ve seen, Feser contends that because the correct interpretation of a rule invariably requires contextual knowledge, any attempt to infer that a purported message is in fact the product of an intelligent agent, apart from all context, is doomed to failure. But what Feser is assuming here is that the identification of a purported message as the work of an intelligent agent requires a correct interpretation of that message. As an Intelligent Design advocate, I disagree: all it requires is the decoding of that message, and it may not even require that. (If the message could be independently shown to be both highly specific and astronomically improbable, I believe it would be rational to infer on these grounds alone that an intelligent agent was most likely responsible for producing the alleged message, even if we had no idea what it was about.) Hence Professor Feser’s assertion that “we just can’t know the right interpretation apart from all context” is beside the point.
Decoding a message is very easy, if it is written in the script of a language we already understand: all we need to do is read each word of the script and confirm that it conforms to the grammatical and spelling rules of the language in question. Depending on the language in question, the code we use when reading the words – something we all learned to do at school – may be either a phonic code (for alphabetic scripts), a syllabic code, a logographic code (for ideograms) or a pictographic code. Even if sentence turns out to be grammatically correct, but semantically nonsensical, like Noam Chomsky’s “Colorless green ideas sleep furiously”, decoding it is still a relatively straightforward affair. And if we found such a message inscribed on the walls of every human cell, we should have no hesitation in concluding that some intelligent agent was responsible, even if we didn’t know who that agent was.
(Note: I should like to make it clear that I do not regard people’s ability to read texts written in their own native language as part of the context of a purported message in that language. Defining “context” in this way would make the term absurdly broad. Rather, I would see the ability to read a language as a presupposition of there being any messages in that language at all. The term “context” refers to circumstances that help us understand the meaning of a message, and does not include the ability to decode a script.)
Decoding a message is harder when it is written in a language we understand, but where the message is encrypted, using a cipher. In such cases, we might think that at least some background knowledge was essential, in order to decode the message. However, there have been occasions when ciphers were reconstructed through the power of pure deduction – for example, the German Lorenz cipher and the Japanese Purple code. Having successfully decoded the message, it would be the very height of irrationality not to ascribe the message to an intelligent agent, even if we knew nothing of the message’s context. For instance, the message might say, “The weather is sunny,” but in spy-talk that might really mean: “The coast is clear: we can proceed with our plan.” But even if we had no idea of the message’s true import, we could still legitimately infer that it originated from an intelligent source, once we had decoded it.
When the message is written in an unknown language, decoding is complicated by the mathematical fact that there’s always some cipher that can be used to transform an unknown message into any string of English characters you want. This point was made by one of my critics, named Scott, who argued: “100 characters of perfectly grammatical English wouldn’t look like any such thing to anyone who didn’t already read English. For that matter, given a hundred of anything, there’s some cipher according to which the series encodes any 100-character string you care to choose.” In practice, successful decoding of scripts in unknown languages, such as Linear A (used in Crete over 3,000 years ago), relies heavily on context-related clues. The question then arises: what should we conclude if astronauts found what appeared to be an inscription in an unknown language on the Moon or Mars? Without a context of any sort, could we still make the inference that the inscription came from an intelligent source?
I believe we can. A simple illustration will suffice. In 2013, two scientists writing in the journal Icarus argued that there were patterns in the genetic code of living organisms that were highly statistically significant, with features indicative of intelligence which were inconsistent with any known natural process. (The authors of the paper, Vladimir I. Cherbak of al-Farabi Kazakh National University of Kazakhstan, and Maxim A. Makukov of the Fesenkov Astrophysical Institute, list several categories of natural processes, and they are clearly familiar with the relevant scientific literature on the subject.) “Simple arrangements of the code reveal an ensemble of arithmetical and ideographical patterns of symbolic language,” they wrote. These features included decimal notation, logical transformation and the abstract symbol zero. Summing up, the authors argued:
In total, not only the signal itself reveals intelligent-like features – strict nucleon equalities, their decimal notation, logical transformation accompanying the equalities, the symbol of zero and semantic symmetries, but the very method of its extraction involved abstract operations – consideration of idealized (free and unmodified) molecules, distinction between their blocks and chains, the activation key, contraction and decomposition of codons. We find that taken together all these aspects point at artificial nature of the patterns.
The authors tentatively concluded that the decimal system in the genetic code “was invented outside the Solar System already several billions (sic) years ago.” (H/t: Max for correction to my wording.)
Regardless of whether the authors’ claims turn out to be true or not – and I’m not holding my breath – the point is that the identification of the signal they claimed to find in our genetic code was made on purely mathematical grounds, apart from all considerations of context. In order to rule out a natural (as opposed to artificial) source for the message, the only thing the authors needed to ascertain was whether it could be accounted for by known natural causes. One could always hypothesize the existence of a natural cause capable of generating these mathematical features, but the authors argue that the only reasonable inference to draw is that the signal they claim to find in the genetic code is an artificial one, generated by an intelligent source.
(I should point out here that our knowledge of what natural processes are capable of generating is not contextual knowledge, but scientific knowledge. As I stated above, the term “context” properly refers to circumstances that help us understand the meaning of a message. Our knowledge of processes occurring in Nature does not help us to do that.)
I conclude, then, that Professor Feser’s contention that the identification of a purported message as the product of an intelligent source cannot be made, apart from all context, is baseless and incorrect. I hope that Professor Feser will be gracious enough to acknowledge this in the future.