Objections to Daniel Dennett’s informational meme.

By Sylvain Magne - 2015


Objections to Daniel Dennett’s informational meme.

1 - Why does Dennett define memes as information?

2 - What is information for Dennett?

3 - Does information make for a good unit of selection?

4 - Can information be transferred without codes?

5 - Does information qualify as a replicator?

6 - So what is information?

7 - Is a word a meme?

8 - Copying, Transcoding and Causal Chains

9 - Algorithms and mechanisms.

10 - Abstractness, information and illusions.

11 - Is Dennett’s informational model viable for memetics?

The concept of information seems to agree with the meme idea, and that is why many memeticists equate memes with information. This view is currently championed by Daniel Dennett himself and it is his own arguments that I want to scrutinise here. I myself also assumed that describing memes as information was a fair bet or at least that it would not hurt the meme idea. I came to discover how Dennett insists on describing memes (and genes) as information, to the point where this would be the only right way of describing memes, as opposed to codes for example. This got me thinking. Considering that I define memes as codes myself (link), I was compelled to try and find out whether the concept of information is a better way to describe memes or not. This is my attempt to better understand Dennett’s information and to find out whether information is really fit to serve as a model for memes. Let's start by recalling Dennett's position.

1 - Why does Dennett define memes as information?

If I get his point right, Dennett is attempting to answer an important question. The main problem for memetics still today is to be able to locate precisely the very subject of its study, the memes. Intuitively, many memeticists, including Richard Dawkins himself, have assumed that memes, if anywhere, should be inside our brains. Dennett himself thinks that memes could be inside brains, while still considering that they may also be outside brains.

Others (including myself ) have argued that it is not possible to have brain memes because it conflicts with the definition of the replicator. Indeed replicators have to be copied faithfully and Dennett himself is aware of that. Dennett knows too well that there can be no physical patterns inside brains that can actually be copied via common human interaction.

Dennett - The New Replicators - 2002 - p5

It is vanishingly unlikely [...] that the brain of Benjamin Franklin, who invented bifocals, and the brains of those of us who wear them should “spell” the idea of bifocals in a common brain-code.

This problem is important for memetics and ought to be resolved. I have offered my own answer to this question (link) but here I want to explore Dennett’s answer. To solve this problem, Dennett suggests that memes are informational abstract entities which have a substrate neutral property. This abstractness would do away with having physical memes in brains, and have informational memes instead which would be independent of their physical reality. Therefore the stuff being copied would not be the patterns inside our brains but instead the information that those patterns code for. Shifting memes from physical entities (or codes) to information would seem to conveniently bypass the issue and fit the definition of the replicator.

Is this a job done? Has Dennett resolved the “memes in brains” issue and found the true nature of memes? Can we now safely claim that memes are informational entities and may reside inside our brains? I would like to raise a few points, put the finger on possible issues with this view, and also point towards alternatives. I will try my best to present Dennett’s ideas faithfully in order to make constructive comments. Please bear in mind that Dennett is still working on these issues himself and may change his mind regarding some of the points, and offer further arguments in the future.

2 - What is information for Dennett?

Most commonly information is understood as a meaningful piece of knowledge, the kind of thing we get from newspapers, teachers or spies. That kind of information has meaning but also, because one piece of information can mean very different things to different people, its content is relative to one’s personal background and interpretation.

A more mathematical definition comes from information theory. In this case, information is the entropy of a particular code. It is unrelated to the meaning of the code and merely measures one single property of that code. This entropic information is not really information as we usually understand it, it is more like a measure of the complexity or unpredictability of the coding of a message.

However, Dennett’s view of information is actually different from those. Dennett has repeatedly referred to memes as being “informational entities” and the following is how Dennett briefly defines information:

Dennett - Intuition Pumps - 2013 - chapter 51 - Herb, Alice, and Hal, the baby

[Information] is an abstract [...] thing that can hop from medium to medium.

Dennett's definition is not mathematical in the way information theory is, however Dennett’s information content is absolute and not relative to people's background in the way that everyday information is. To Dennett, information is a conceptual object which can materialise on different types of media, and therefore appear as if it could hop from medium to medium. The very same information could be carried by images, or words or computer chips and remain unchanged and absolute, independent of those media. Information would transcend matter. Dennett has also labelled this property “substrate neutrality”.

On the other hand, Dennett also acknowledges that there is nonetheless always a need for a physical medium when he says:

Dennett - Intuition Pumps - 2013 - chapter 51 - Herb, Alice, and Hal, the baby

Of course any actual copy of [information] has to be made out of something (if not ink, then maybe letter-shaped patterns on a computer screen, or even strings of binary code burned into a CD), [...]

This said, there is one thing that Dennett thinks is not necessary for information transmission, and that is the concept of code. I want take a closer look at those concepts of information, codes, abstractness, substrate neutrality, etc. and see if this informational model is viable.

3 - Does information make for a good unit of selection?

To Dennett, information is substrate neutral, and as a result, the exact same “word-meme” is found in its spoken and written versions, but also in its English and Italian versions, in its brain and computer versions and all its possible versions.

Dennett - The New Replicators - 2002 - p6

A recipe for chocolate cake, whether written in English in ink on paper, or spoken in Italian on videotape, or stored in a diagrammatic data structure on a computer's hard disk, can be preserved, transmitted, translated, and copied.

Unfortunately, because of this substrate neutrality, this informational model alone does not allow us to differentiate between all of these versions and therefore ignores the differential success of English vs Italian or spoken language vs written language. It suddenly wouldn’t matter if entire languages disappeared. Memetics would ignore what makes a synonym more successful than an another, or what makes written latin more successful than spoken latin. To be fair, Dennett concedes the issue to some extent but downplays the importance of it.

Dennett - The New Replicators - 2002 - p7

It is possible [...] that encodings of the same meme in different media will differentially compete and differentially mutate, so that they should be considered different memes for some purposes.

Yet any useful evolutionary model of culture should account for those differences, and not just relegate them to minor exceptional cases. These differences are not the exception, they are the rule.

On the other hand, if we were to consider memes as being codes instead of information then this would allow us, at least, to take into account the differential rates mentioned, as written and spoken languages are very different coding types indeed. Every translated word in every language, and their synonyms could be considered as competing for the same place in the meme pool.

4 - Can information be transferred without codes?

Dennett asserts that information can be passed on without any encoding.

Dennett - In response to an email I sent in 2015, Dennett replied this:

“Codes” isn’t quite right, I argue; you can have transmission of a meme without any ENCODING. (my old example of a wagon with spoked wheels that transmits the meme of a wagon with spoked wheels is a case in point). I’m not giving up such examples since I think they abound, without codes.

Dennett - Darwin's Dangerous Idea - 1995 - p348

A wagon with spoked wheels carries not only grain or freight from place to place; it carries the brilliant idea of a wagon with spoked wheels from mind to mind.

In other words, the argument is that if memes can be transmitted without using codes then it is likely that the concept of “code” is not the best suited for describing memes. Fair point, and as a matter of fact, if we could find only one single meme that could be transmitted without encoding, it would be enough to invalidate my claim that memes are codes.

However, I argue that there is encoding indeed, not only that but there has to be encoding. Going back to the wheel, I hope that Dennett would agree that if an engineer drew a plan of the wheel, then the plan would be a code. Now, the engineer may happen to be a very good artist and he/she may be able to render a lifelike painting of the wheel. If so, then what would be the difference between the lifelike illustration of the wheel and the wheel itself? There would be none. Both objects would broadcast a code made of geometric shapes that our expert pattern-sensitive eyes and brains could see and recognise. The light bouncing off both the engineer’s plan and the wheel, is not random, it displays patterns of photons. As a matter of fact, codes and patterns are quite simply the same thing, they are synonyms. So yes there are codes indeed. In this case, the code takes the shape of light wave patterns.

Furthermore, on a more general, if not universal, note, it is logically impossible to communicate information, or anything, without encoding. If there are no patterns, there is only noise and if there is only noise, there is no information. In order to communicate, there is a need for some pattern to be organised in some way, and that pattern is the code.

5 - Does information qualify as a replicator?

If replicators were made of information, then it may imply that it is specifically the information that is being copied, and not codes. If so, how would we go about proving that information is being copied and being copied alone?

Let’s take an example. How would we go about determining if two strands of DNA carry the same gene or if two computer files carry the same instructions? The way we would do that, in practice, is usually by looking for similarities in the codes. In the case of genes, the code is in the patterns of nucleotide sequences, so we would compare the molecular sequences in both copies and make sure they match. In the case of computer files, the code is in the sequence of digital bits and again we can establish whether they are identical or not by comparing the codes.

When looking at communication closely, one can observe that it is never information that is copied, but it is codes that are copied (or transcoded). We do not look at information directly because how would we do that other than by looking at the code? There are no tools to actually evaluate information. There are only tools to evaluate codes.

This simple fact undermines the very existence of information, or at least the way Dennett sees it. More importantly, it makes information a doubtful candidate for being a replicator, because replicators are supposed to be the very stuff that is copied, as Richard Dawkins explained.

Richard Dawkins - The Extended Phenotype - 1982 - p83

I define a replicator as anything in the universe of which copies are made.

6 - So what is information?

Dennett describes information as abstract or substrate neutral.Those two concepts seem very closely related in Dennett’s view. Unfortunately both concepts are rather vague and open to interpretation, but of the two, Dennett offers a more precise definition for substrate neutrality. Dennett sees similarities between information and algorithms, and uses the algorithm for long division to explain substrate neutrality.

Dennett - Darwin's Dangerous Idea - 1995 - p50

substrate neutrality: The procedure for long division works equally well with pencil or pen, paper or parchment, neon lights or skywriting, using any symbol system you like. The power of the procedure is due to its logical structure, not the causal powers of the materials used in the instantiation, just so long as those causal powers permit the prescribed steps to be followed exactly.

According to Dennett information is substrate neutral in a similar way to algorithms. To Dennett there is something real there which is not just magic or just an idea in the brain, it is something else. This example seems quite compelling but I want to point out that Dennett’s argument is not really an argument but essentially an analogy between information and algorithms. Ultimately, this analogy does not explain how algorithms manage to hop (if they do) and what it means really to be substrate neutral in plain physical terms. Dennett also often mentions how words are abstract and very real at the same time. This argument is also an analogy between information and words and does not give a clear insight on what information is actually supposed to be.

I fully admit that it is undeniable, to any observer, that there is something in common between a computer and a person when they are both performing a long division, even though they go about it in very different ways. I fully admit that something appears similar between a spoken word and a written word. However giving this similarity a name does not explain it. Calling it substrate neutral or abstract is not enough to make it a subject of scientific study. I am afraid Dennett’s take on information lacks definition and precision and I believe there are better ways to describe information.

I suggest a change of perspective.

In my view, information is actually not on the replicator’s side but on the phenotype side. Let me explain. Just like it may be convenient to say that long necks have spread among giraffes when we know that (excuse the shortcut) it is the “genes for long necks” that have spread, it may also be convenient to say that information spreads when in fact it is codes that spread. When we communicate, we don’t actually really send out information, but instead we broadcast codes. Those codes are a strategy to build meaning in other people’s minds. At the scale of a population, it appears “as if” information can spread from brains to brains, but in fact information is not communicated, just like long necks are not communicated. Information is built, a bit like an organ is, but in the case of information we are talking about brain programmes. Indeed, information would be better described as interpretation or meaning or a brain programme because information happens in the brain only.

Here’s an analogy. Let’s imagine that a war is going on and a general needs a new canon to win the battle. In case it is too difficult to carry the canon, too costly or too slow, it may be faster and cheaper to just send an engineer’s plan and build a new canon in situ. Once the new canon is built, it may just be “as if” an actual canon had been sent to the general. However we know very well that it isn’t the case. Also, it may happen that the new canon may not be quite the same as the original canon because the materials and tools used on site may be different. In this sense the canon is very similar to the information that we get from reading a newspaper. The canon may look different on each battlefield just like the news may be interpreted differently in every reader’s mind.

Information is the result of the interaction of a code and a reader.  It is only because all human brains are similar readers (to some extent) that we expect a piece of code to look “as if” it were literally “carrying” a piece of information.

If Dennett were right and we were actually sending information then people would receive that theoretical abstract information intact. There would no room for confusion or interpretation. But that’s not what happens. People can be confused and misunderstand a message. That’s because they use the code received and their own knowledge to build their own information, their own interpretation. For example, 911 can mean a “terrorist attack”, a “Porsche car”, an “emergency phone number” and it can also mean nothing or anything. This is how the environment has an effect on the expression of a memetic code, just like the environment has an effect on the expression of a genetic code.

As a result, information is not strictly substrate neutral because it only has one substrate and that is the brain it resides in (or computer brain, or the cell). However, information could be said to be code neutral in the sense that a brain may build the same information from different types of coding such as text or speech. Just like we could learn a song from a text or a recording, or we could bake a cake from them. In that sense information differs from organs because organs are only built from one type of code (DNA-RNA), whereas brain programmes can be built from different types of codes, (sounds, light, textures, etc.)

As human beings we care more about the meaning of a message rather than the nature of a message. When we speak, we don’t consciously craft sound waves. Even less so when we hear a sound. We’re not even conscious that the sound goes through our ears. On the other hand, when we write we are a lot more conscious of the codes that we are using because we’ve had to make a conscious effort to learn them. But when we communicate we don’t really care how we send a message, orally or in writing, what we want is to make sure the person receiving our message will build the same information from it. On average it is fair to assume that most people get the same information from the same code, yet as we have seen, it is not necessarily the case.

Because we care about the meaning more than the codes, we tend to ignore the relevance of the code, and I think that’s where Dennett gets it wrong. The concept of information may be a good approximation in some cases but it is ultimately wrong.

Understanding meaning is crucial to understand the fitness of memes. Brains do not exercise a selective pressure on codes by selecting the codes directly, they do it by validating the meaning that they build from the code. We do not see light waves or sound waves, we only “see” the meaning of them, which is the interpretation that the brain creates. When we hear the word tree. We do not imagine sequences of air pressure, certainly not, we have a personal experience of the word tree. That personal experience is the meaning. If we validate this meaning, we indirectly validate the code that caused it. This is like selecting genes by their vehicles. The vehicle of the meme is the brain programme that it created, while the meme itself is the codes that came in contact with us.

Should we get rid of the concept of information? Maybe not. Here’s a thought. Instead of talking about codes carrying information, we could talk about codes having an informational potential, a statistical ability to cause meaning. In other words a code could be clear or vague, open to interpretation or not, widely recognisable or not, etc. For example the word “taxi” could have a higher informational potetial than the word “abstract”. That’s because “taxi” is more popular and better defined than the word “abstract”. Yet, a written “taxi” and a spoken “taxi” may not have the same informational potential. That’s because the sound of “taxi” may be universally understood while it may spell differently in different languages. Each meme could have, depending on its environment, a different informational potential which influences its potential for survival. I would be more satisfied with such a definition.

7 - Is a word a meme?

Dennett often argues that memes exist because words are memes and words exist.

Dennett - Breaking The Spell - 2006 - p80

Do memes exist? Yes, because words exist, and words are memes that can be pronounced.

However, this argument is borderline circular. To any person willing to know what a meme is, the statement “words are memes” is no more obvious than the statement "memes exist” and either of those call for just as much explaining. That aside, I only agree partly with the statement that words are memes. Unlike Dennett I think that words are not accurately described as just abstract entities.

Dennett - Breaking the Spell - 2006 - p80

Words are such familiar items in our language-drenched world that we tend to think of them as if they were unproblematically tangible things—as real as tea cups and raindrops—but they are in fact quite abstract, even more abstract than voices or songs or haircuts or opportunities (and what are they made of?). What are words? Words are basically information packets of some sort,[...]

To me, words are indeed very real and truly made of many things, and I think they can actually be compared to organisms. Words are similar to organisms because they have a kind of genotype (code parts) and a kind of phenotype (body parts). Their genotype is made of memes of course, which can be of several possible types such as sound memes, light memes, texture memes (braille), etc. while their bodies are made essentially of brain programmes (or computer programmes) and prints of many kinds. The bodies of those words have a sedentary existence inside our brains (computers and prints), while their memes travel between those meme machines. Note how those different memes, even if they build the same brain content, they may compete against each other a bit like alleles do. We are, to some extent, made of words, and programmed by memes.

8 - Copying, Transcoding and Causal Chains

Instead of talking about information, could we describe how communication and replication work in plain physical terms? How are these memetic patterns passed on exactly? Is there some informational ghost or informational black box that is being secretly passed on without us noticing? Unlikely. At the end of the day it’s all just physics, right? So let’s try and describe the transfer of patterns without calling for abstract entities.

It all starts with the observation that, in nature and in culture, some patterns are recurring. What we are trying to explain is how those patterns get to reappear. What are the mechanics behind it? In principle, we could describe replication simply as a sequence of events which starts with one pattern and ends with a copy of that pattern. For example, let’s say we start with a pattern A, then somehow A becomes another pattern B, and then B becomes pattern A again.

The question is, what do we need to ensure that A is recreated faithfully?

We understand that this transformation process needs to allow for A to be recreated at some point down the line. For that to happen, when A changes into B, the pattern A needs to be transcoded into a new pattern B, bit by bit. For example, that’s what happens when DNA is transcoded into RNA, or when a wifi signal is transcoded into an electric signal. To each bit of DNA pattern A there is a corresponding bit in RNA pattern B. This allows ultimately to revert the process and regenerate A from B. The fact that A and B are mutually transcodable ensures the possibility to recreate A. To allow copying, each step of the sequence of events has to be mutually transcodable or reversible.

Each step will follow some mechanical rule of transcoding, or another. Much transcoding can happen before it goes back to the original pattern. This series of events, put together, form what we could call a causal chain, simply because there is a causal link between each step of the chain and between each bits of the patterns.

An example of a successful copying process is when someone says a word, a second person hears it and writes it down and a third person reads it out loud again. This causal chain went from sound wave patterns to brain patterns, then to ink on paper patterns, then to light wave patterns, then to brain patterns again and finally to sound wave patterns again. Apart from the first one and the last one, none of these patterns are copies of each other, they are only transcodes of each other. There’s an important nuance here.

Indeed, transcoding alone is not copying. Two memes can be transcoded into one another and yet remain two different memes. For example, if we transcode the English word “tree” into the French word “arbre” we get two different words. As a result, “arbre” is not a copy of “tree”. However, transcoding can be used as a tool to create copies. Thus it is possible to transcode “arbre” back into “tree” and by doing so we create a new copy of “tree”.

That's also why written and spoken words are not the same. The spoken and written languages are in effect two different languages coexisting side by side. We all speak a sound language and a sort of sign language called writing. There are strong correlations between the two but there are also differences.

Causal chains are neither rigid nor unique and the path they take can be varied and complex to say the least. Those causal chains can run through our brains, through artefacts, through behaviours, through light waves and sound waves, through textures, through computer brains, and more. They can cross and combine in many ways via some kinds of natural or artificial boolean logic gates, they can also break or be repaired, they can split or merge. The specificity of evolution is that it has causal chains that are reversible and allow copying. A reversible causal chain can keep going and looping for centuries, it has the potential to be eternal.

For memetics, the question is to find out which link in the causal chain is the actual replicator. Are all links replicators? Is there only one replicating link or can there be several? To put it simply, the links that are likely to be replicators are the links that never change. It is the case because, as we know, replicators are meant to be copied faithfully. On the other hand, the links which do change are likely going through either the environment or through the phenotypes (the meme machines). So for example, a spoken word will, most of the time, be spoken in the same sound wave patterns, but when it is transcoded into brain codes, these brain codes will be unique to each individual. Therefore, the replicator is more likely to be the sound wave rather than the brain code. A causal chain can involve several memes, like when for example a word is both written and spoken in succession. The spoken and written word compete but they also cooperate in replicating each other.

Causal chains, if we could see them, would form a complex intertwined web. This web of causal chains is the tree of life of memes, and it looks very different from its biological counterpart. In my view, in order to explain what happens when memes are copied, rather than talking about some abstract and elusive information being passed on somehow, talking in terms of causal chains is much more useful and enlightening. Causal chains are the lifeline of memes.

9 - Algorithms and mechanisms.

Dennett describes algorithms as abstract entities but there is a way to describe them in a more practical and useful way. To fully understand what algorithms are we also need to understand what mechanisms are because these two are inseparable.

Algorithms and mechanisms are two very similar concepts which could easily be interchanged but there is an important difference between them.

Let’s start with mechanisms. A mechanism is a set of physical parts that can perform a particular task. Any physical phenomenon, such as tides, winds, chemical reactions, movements of planets, etc. can be interpreted as mechanisms. However, not all mechanisms are interesting and that's why we usually keep such terms to describe the more useful of them, or the ones that we create ourselves for a purpose. Indeed we design mechanisms to solve problems. Common examples of mechanisms are clocks, washing machines, bicycles, typewriters, wagons with spoked wheels, etc. Even inanimate objects such as a chair, a book or a photograph are mechanisms because they perform a task such as supporting a person at a certain height or displaying written characters or images. Computers are also mechanisms even if their mechanical parts include small particles such as moving electrons. A computer is a very interesting mechanism because it is capable of creating new mechanisms called programmes. Dennett likes to call programme virtual machines, which is quite fitting, but virtual doesn’t mean that they are “not real”. If we look at how programmes work, they are essentially complex networks of microchips and moving electrons.

Our bodies are also mechanisms, albeit very complex ones. We are made up of cellular factories of extreme complexity and in very large numbers but yet, it is all mechanisms. Our brains, much like computers, can also create mechanisms. Those brain mechanisms are to brains what computer programmes are to computers.

Let’s turn to algorithms. When I said that a mechanism is a set of physical parts that can perform a particular task, the algorithm is simply a description of that task. Indeed, I could have written that a mechanism is a set of physical parts that can perform a particular algorithm. An algorithm is a kind of theoretical model for a mechanism, it describes what a mechanism could do but with no necessary concern about how the mechanism is built exactly. A useful algorithm is meant to give clear instructions about each step of the task so that the task can be performed accurately. For example, one could write an algorithm, in plain english, describing how to sort coins. It could read:

  1. Allocate a container for each type of coin
  2. Take a batch of unsorted coins
  3. Take one coin out of the batch and place it in the corresponding container
  4. Repeat step 3 until the batch contains no more coins

This algorithm could then be implemented into a mechanism able to sort coins. This is exactly what coin counters do, which are the kind of device you find in a vending machine for example.

There are many different ways to build a coin counter, and indeed there can be many different mechanisms that can perform the same algorithm. Dennett’s favourite example is the long divisions. The procedure for long divisions can be expressed as an algorithm and it can be implemented and performed by an automaton, a computer or a person. In fact, for one single algorithm there can be an infinity of mechanisms. In other words there are an infinite amount of ways to perform the same task.

An algorithm is not unique either, because there are also many ways to compose an algorithm. For one it can be written in different languages, but also in different computer languages. Even within the same language, it can be formulated or phrased in different ways. The whole of mathematics can be interpreted as being a huge set of algorithms and in mathematics as well, one can write two equivalent algorithms in two different ways, the same way that one can make a mathematical demonstration in different ways. Unlike what Dennett thinks, all of these different versions of an algorithm are not really one and the same algorithm, they are simply equivalent to one another. This difference is important because some algorithms will be more convenient to use than others, and in that sense they remain different. Very much so from the perspective of an evolutionary model.

On one hand, the concept of algorithm is interesting because it shows that there is an equivalence between different mechanisms and on the other hand, the concept of mechanism is equally interesting because it shows that there is an equivalence between algorithms. This relation between the two emphasises the fact that a task can be equally performed in many different ways by different mechanisms and equally programmed in many different ways by different algorithms.

This explains why there can be a selective pressure on both algorithms and mechanisms. Because there are many ways to do the same thing, some ways are bound to be better than others and more likely to be replicated. Indeed, to us humans, it will make sense to try and create the most efficient and cost effective algorithms and mechanisms, in order to help us solve our problems.

What are mechanisms and algorithms from a memetic point of view?

With regard to algorithms, from the perspective of memetics, an algorithm is quite simply a code. In its largest sense, the concept of code and the concept of algorithms are actually identical. This means, for example, that all replicators are algorithms, however, not all algorithms are replicators just like some codes are not replicators.

With regard to mechanisms, from the perspective of memetics, a mechanism is synonym to phenotype. Where there are replicators that build machines, these machines are both mechanisms and phenotypes.

So we can look at memetics in terms of algorithms and mechanisms in just the same way that we look at memetics in terms of codes and phenotypes. Memes are algorithms that can programme brain mechanisms and computer mechanisms.

10 - Abstractness, information and illusions.

Dennett struggles to give any clear definition of abstract entities other than a few analogies. I believe one can be a lot more specific about their nature.

Abstract entities are simply ideas inside our brains. Ideas, concepts and mathematics have been called abstract because they seemed detached from the real world. They seemed real and yet not made of anything. This was in a time when people may have had very little idea of what a brain is and what it does. Now we can see that ideas don’t need to live in an ethereal world of ideas but exist within the complex network of neurons that make up our brains. Just like computer programmes exist inside computers, all ideas and abstract entities are brain programmes existing inside our brains.

Those ideas never leave our brains because only codes can travel between brains.

The same goes for information. Information is just an idea, a brain programme. The continuity that Dennett sees with information, or substrate neutrality, is an illusion. We may get a strong feeling that a spoken “horse” and a written “horse” are the same but it's only because we can link these two concepts in our minds. Think about how we see the same thing in time and money when we say “time is money”. Obviously time is not money but we can project mentally how we can use time to make money or spend money to save time. The apparent continuity that Dennett sees in information is simply due to the fact that our brains see similarities and project those similarities back onto the real world. The fact that our brains make virtual links between things does not make them the same. The continuity of information is nothing more than an illusion, it may be a convenient shortcut but it does not correspond to a real phenomenon.

What else could we say about ideas / brain programmes?

There is a lot more to say about abstract entities when we understand that they are simply ideas.

11 - Is Dennett’s informational model viable for memetics?

If my objections are valid, then I am afraid not. I think that the informational model is rather compelling at first glance but when looking at it closely, it seems to break down. Information is still a vague concept that does not lend itself well to empirical study and it seems not convincing as a unit of selection or as a replicator. Dennett doesn't actually offer a clear perspective of what his model could bring to the table and what could be achieved. Also, his model is rather sketchy to say the least. Apart from a few analogies there is not much substance to this model. Dennett’s definitions are few, short and rather vague. Nothing that could be used to build a science without doing a lot more work to it. I fear that the informational meme has little to offer and would make for a rather fruitless theoretical basis for a science of memetics.

For these reasons, I still claim that the concept of code is a better model for memetics, and here’s why briefly:

I am glad I have taken a closer look at Dennett’s model because it pushed me to take my reflections further. I am now even more confident that my code model is a much better model for memetics. You can find the core principles of my model in my previous articles about the code view of memes: link

This said, there is much more that needs to be done. Even if a theoretical model is accurate, it can be useless if it doesn’t make any good predictions. I have no doubt that this can be done with the code model but being the only person I know working in that direction means progress is slow. I have come to realise that one of my challenges will be to get people to simply read my work. Time is precious for everyone and few are the people willing to give me some of their own time. I will continue nonetheless because it is worth it, and memetics deserves to be developed.