Knowledge Sharing – A Critical View

Memory Specialisation

Memory specialisations are handled by functional processing areas. That is, the same area that processes information from senses or from reflection, also deals with the memory associated with that specific feature of a scene.

Collective Memories

However, when we utilise memory, we tend to do it as a collective scene and tend not to separate memory into its individual features. In some cases it may even be difficult for us to isolate the features of a specific memory even though they are stored in separate parts of the brain. For instance, if we were to recall an instance of a garden party that was attended some time ago, it may be quite easy to remember things from the event but it may be more difficult to remember only an isolated feature of the event such as just the sounds or only the smells or just the touch sensations from the event, in isolation from the other features.

Features can trigger recall

However, any of the features of that memory could trigger the memory at some later time. For instance, a specific smell could cause the memory of the garden party to return even though it was not the main feature of the memory.

Human memories are content addressable

Search engines allow us to look through stored information using some key to its content. We don’t need to know the address of a document to search for it using a search engine. Knowing the address would of course, imply that we didn’t need to search, because we would know where it was, we would simply need to retrieve.

Content addressable as long as it’s text

Unfortunately, this computer based form of content addressability only works with written text and not with sounds, images, video, touch or odour. In most cases though, the only way that we can initiate a search is by entering text. It is more difficult to enter a smell to search for even if it was possible to do so. Even entering an image to search for is not that simple. One could select and image an then ask the computer to find similar images, if we could specify what similar is to mean in each search instance.

Humans use scenes to find memories

When humans find memories, they may also be expected to locate the memory from something that is or could be written text. For instance, do you remember what happened at the garden party we went to last June? However, the human is likely to quickly revert to a scene in order to answer the question and would not search a text based report of the events that took place at the garden party.

Computer searches are still very restrictive

Since we find it difficult to express, in words, many of our sensations, we can find ourselves reverting either to analogy or the identification of example or similarity. For instance, what does this smell remind you of? Or, do you remember that car we saw on holiday that looked like a frog. Unfortunately, searches don’t even allow us to look for pictures of cars that look like frogs, unless someone has labelled an image or described an image using text, that provides information that is searched instead of the image itself.

So called sharing involves a lot of translation

Yet, most of what we, probably mistakenly, call knowledge sharing, involves much more translation from one representation to another than it does sharing. If I am asked to explain something, I am likely to picture the situation or thing and then attempt, in the best way that I can, to describe it to the person asking for the explanation. When they hear the explanation, then they attempt to rebuild the scene that I am describing, using their own experiences and background knowledge in order to do this. The thing that gets understood by the receiver, may not be exactly what is being pictured by the sender. Often, the meaning of terminology used by someone with one disciplinary background may not match the meaning of terminology used by someone else with a different disciplinary background. When such terminology is simply used as a small part of a larger explanation, both parties involved may be happy to accept the term without question and may never realise that it means different things to each of them.

Other people can’t access our memories directly

Human memory is not like information stored on a computer, there is no common access to it. Only the individual that possesses the memory can access it directly. Anyone else can only get an explanation of it from the owner and the owner must translate from the representation that his or her brain is using for that memory into spoken or written language.

Sharing common representations

One way that people have addressed this difficulty, and it is a difficulty, is to create common models and representations for information that people can interact with directly. In this way, each person must interpret the information from the model but does not need to interpret his or her own memory in order to explain something to someone else. The person can refer directly to the model or representation when discussing his or her knowledge. In a similar fashion, if a person has added to or changed the knowledge, then he or she will then make changes to the common model, to reflect the new. Others are then in a better position to understand the new because it is placed within the context of the common model and does not require a new explanation.

Mathematics is the best model

Probably the cleanest and most unambiguous way of creating common models is through mathematics. This is so because other people have created a rigorous modelling language in mathematics that if appropriate, can be used as a common representation for an individual’s knowledge that others can share and challenge objectively.

Diagrams can work well as representations

Diagrams can become another such model for a person’s knowledge in some cases. This is more justifiable (probably only justifiable) if the diagramming method has been properly established and accepted before it is used to model a person’s knowledge. Flow charts are a good example of a diagrammatical model that can represent some aspect of a person’s knowledge and be used unambiguously by others that understand the diagramming method, to acquire or use this knowledge. There are many other examples of methods of representation involving diagrams; some of these also employ mathematics within the model.

Redundancy in conversation can help address the problem

Without some fairly rigorous form of common representation, it may be difficult to justifiably claim knowledge is being shared. This is because of the way that humans actually represent knowledge and the need for multiple translations during conversation. However, there are some elements of conversation that can take advantage of human representations and at least partially facilitate the transfer of knowledge from one person to another. In order to achieve this, at least in part, there must, out of necessity, be a lot of redundancy in the information transmitted and received. This redundancy is used to help to paint pictures inside the participants of a conversation. There can be two pictures involved in this painting process, the one that is supposed to represent the knowledge itself and another that can represent what one person thinks the other can see. Without the second view, it would be difficult for the original holder of the knowledge to adjust the conversation to correct any misconceptions and misunderstandings that may be made by the receiver. Likewise, the receiver needs to be able to understand what the first person is trying to get at and then be able to explain where problems are being experienced.

Models need to model correctly

This method has worked well for humans throughout the history of civilisation but may not always be good enough to represent knowledge of a scientific and very precise nature. In such cases, models are needed, and where finite precision is required, these are usually mathematical models. There is of course a requirement that the models can actually model the knowledge correctly and fully.

Sharing or Translation

So knowledge sharing may be possible through a common model, otherwise it is translation that is involved rather than sharing.