• robgillespie


    I proffer my head to the lion: provoking debate is the path to enlightenment. Musings on micro-content.

    Content molecules are free-form and volatile content elements that have valence. They are continuously updated and can be dynamically formed and chained. They are offered in response to triggering events defined as request contexts.

    The paradigm has shifted from information being determined by the supply-side to it being determined by the demand-side.

    Content molecules are composites of smaller components, molecules or sub-molecules. They can contain a variety of content types, for instance, text, VR media, graphics and so on. Different content types fulfilling the same purpose enables many contexts to be supported.

    Sub-molecules are components of molecules. A molecule may have a single component, but more typically contain a number of components of different types. The defining characteristics are that together they have a (presumptive) title, are conceptually consistent and are a valid response to a request.

    Molecules and so sub-molecules are by definition relatively small, but depending on the subject, purpose, enterprise logic and so on, could contain quite significant amounts of conceptually consistent content.

    The nature, composition, and size of a content molecule are determined by:

    • Purpose: Content must have a clearly defined purpose: it must fulfill a need and have value. It might answer a particular question, such as What is this? or How do I do that? It may also fulfill an enterprise purpose, such as to raise awareness or promote a product. Some molecules and in particular sub-molecules could, of course, fulfill many different purposes in many different combinations.
    • Enterprise logic: Enterprise logic reflects the value of the content to the creating enterprise and the business processes that determine the content that is created and made available.

    • Validity checking: Content must remain valid. Validity checking is critical to maintaining the integrity of the content and the information provided in response to a request. Dedicated intelligent agents and feedback loops will provide ongoing sanity checking. Governance structures for Artificial Intelligence will help maintain the validity of the offered content molecules and molecular chains.

      A content molecule is a standalone unit of micro-content that is:

  • Conceptually consistent: Each molecule and sub-molecule must be conceptually coherent. Conceptual consistency is strongly related to maintaining the relationship between the contents of a (sub-) molecule and its title or presumptive title. I say presumptive because clearly, not all content (or perhaps any) requires a (static) title. The identification of the nature and purpose of content can also be by other means such as metadata or intelligent content analysis and machine learning.
  • The least specialized possible to fulfill its purpose: Appropriate specialization is not a bad thing. Content should be clear accurate and relevant. However, inappropriate specialization prevents reuse and requires the creation of substantially similar content without real delivering additional value.

    Content molecules and their components are only offered; no (sub-) molecule is mandatory. The offer of content is determined by the context. Certain quite obvious and legitimate exceptions must exist. For instance, legal statements and similar.

    Content molecules can be formed and reformed into larger molecular structures on demand. The molecular chain is shaped by the request context.

    Chaining must not be restricted by content creation practices that prevent content (sub-) molecules from standing alone. Obvious contra-indications include stubs, stems and location flags.

    Similarly, requests and contexts can be concatenated. This is particularly so when an information request cannot be resolved with a simple content offer. A valid response would be a request for further information or, a content manifest.

    Content manifests would be provided in response to requests when the potentially available content is voluminous and or varied. The manifest is an offer of content to which the requestor can respond. This implies that a variety of content molecules or molecular chains have been identified as potential responses.

    It seems likely that a library of well-known responses/patterns (to well-known requests) is generated. These patterns can then be applied as is, or in an appropriately modified form.

    Any content offer containing suggestions for related and supporting content can lead to concatenated requests and so contexts.

    The role of metadata is critical. Generally, the use of data about data and more specifically, as a form of transmittable specifications enabling the content to be presented in different forms in different contexts. The latter would presumably be maintained in content objects.

    It is a subject that deserves careful and detailed consideration.

    Feedback loops enable requestors to request that content is created or modified. The loop can be direct, such as a message, and/or indirect through the application of analytics to request-related data. Analyzing and applying such data is a key factor in assessing the validity of content offers and the need for new or modified content.

    Community content is powerful. Users must be able to generate content, and enterprises must be able to incorporate that content in responses to requests. Content creation will be a collaborative endeavor.

    Content is developed by many people (and machines), for many different purposes at different stages of a development cycle. Content silos should not exist because of the inability of Enterprise CMSs to interwork, nor because of limitations imposed by format or file-type. Inhibitions on content sharing or availability must exist only for legitimate business purposes, such as privacy, compliance requirements or the protection of intellectual property.

    In the absence of legitimate inhibitions, at the very least, virtual editing suites should allow content to be reused or modified on demand. The availability of content to meet user requests must not be inhibited by artificial requirements and processes such as recreation in specific formats, or by specific people.

    Content must be created and most importantly made available continuously. The requirement to distinguish between content for a particular release or version would not be a question of compilation into documents sets (since neither static sets nor documents would surely exist), and more a question of the request context being fulfilled.

    Availability logs and delta information will have a particular value, both as a subject for a request and a trigger for defined actions.

    Machine readability must go beyond simple text/video-to- chat capabilities. Moreover, in addition to machine-to-machine communication, there will be complex inter-working between machine-generated requests and human requirements for information. Contexts will surely be complex, with chains of mixed and concatenated requests.

    Business process automation is a given. Such automation will give rise to well-known information request contexts. Such request patterns are likely to be shared between the requestor and the information provider. Furthermore, the provider is likely to generate and maintain information provision patterns (as a service) that are triggered and executed according to defined criteria.

    A simple example would be where a requestor subscribes to updates or changes to content molecules or particular classes of information. The availability of content matching the requestors subscription would trigger a pattern to be retrieved from the pattern database. The pattern would define a set of actions. The pattern would then be executed autonomously.

    Although the same pattern may be used for many requestors/subscribers, each triggering would constitute a new context or set of concatenated contexts.

    Context objects contain behavioral information and context for each molecule. There is an expectation that many molecules might be required to respond to a request, but unlike in a book format, the order, structure, and relationships between them are, at most, minimally defined. It is the requestor, the context, and requestor- profiling that helps determine what comes next.


  • Alex Masycheff


    A great definition of molecular content! I would emphasize a couple of more things:

    1. Because content molecules are offered as a response to a request context, this means that the channel through which the content molecules are offered is also determined by the context. In one case, it might be an online customer portal. In another case, it might be a chatbot. It also might be an augmented reality app where content molecules are rendered and displayed over physical objects.

    This reminds us about the multi-channel publishing concept. Separation of content from presentation lets us publish the same content to Web, PDF, ePub, etc. With content molecules, content is separated not only from the presentation, but from the business logic, that is from the way the content is processed and manipulated. The business logic is determined on the fly based on the context.

    2. The context may also determine the form in which the content is offered. For example, if the user is a maintenance engineer who is looking for information on how to perform scheduled maintenance for a specific component, and the procedure is simple, then it can be displayed as a plain text. But if this is a complicated procedure which includes multiple checks, it could be more convenient for the user to get it as a dynamically generated flowchart representing the information visually. The flowchart is not created manually in advance. Instead, it’s generated automatically if the context requires that.

  • tressaire


    Building on the importance of metadata, molecular content must be properly “tagged” and identified in order for “machines” to use it in proper context. As Alex stated, the content is separated from the presentation, therefore, the proper metadata is key. Remember, despite our use of AI as a terminology, we are not there yet, to quote Andy McDonald. Machines are not yet “intelligent”; at most they are trained to recognize patterns.

  • Yoel StrimlingYoel Strimling


    At the STC Summit in 2008, Michelle Corbin of IBM and I gave a well-attended presentation called “Editing Modular Documentation: Some Best Practices” (http://www.writersua.com/articles/modular/), which was then published in Intercom.
    It addresses a number of the issues brought up in this article, and you might find it interesting.

  • Alex Masycheff


    Agree. What’s interesting is how metadata can be used to train intelligent chatbots, the ones that capture user’s context and match it with content to be offered to the user.

    There are several fundamental concepts in chatbot development and training, and they include intents and entities. Intent is what the user wants. Entities are parameters of the user’s goals. E.g., if the user asks “How can I copy a file to an external storage disk”, the intent might be “copy a file”, and the entity might be “external storage disk”.

    Metadata assigned to content has to reflect the intent and entity. But this is a bi-directional correlation. Based on the metadata, you can better understand intents that your users may have. And the other way around – if after analyzing actual user’s requests, you found an intent that is not reflected in the metadata yet, the metadata has to be updated. This is a loop that in the end of the day helps you effectively train your chatbot.

  • Ray Gallon


    This may be an aside in this thread, but I think it’s important to mention: I don’t care if artificial intelligence is “intelligent” or not. There is deep learning technology, and it is an oversimplification to say it just recognises patterns. It does actually “learn” in the sense that it acquires knowledge about things it didn’t ever know. Pattern matching is a significant part of that, but the algorithms it uses, while not new, are quite sophisticated, and arguing over whether it is intelligent or not misses the point.

    The point is that these technologies are in play, and they will get ever-more efficient and powerful as they develop. They don’t need to be like us, they need to do things that we don’t do well, so we can do what we do best in complement.

    Molecular information is part of the equation designed to help us do just that – and how we manage, work with, and interact with autonomous machines (intelligent or otherwise) will have a lot to do with our success.

  • Viewing 6 posts - 1 through 6 (of 6 total)

    You must be a consortium mmber to reply. Click here to join or sign-in, if your already a member.