Category Archives: Intelligent Content

Ontology for the Perplexed

Sooner or later people who deal with content hear about an odd word called ontology.  It is often discussed as a forbidding topic: the representation of all knowledge, and the source of endless grief for those who dare to wrestle with it.  I’ve seen online debates of people trying to define what it is, often by comparing it to other forms of content organization.  These definitions are sometimes theoretical, sometimes impossibly mechanical, and routinely confusing.

Ontology is often shrouded in mystery.  But it is an important topic with practical uses.  Ontologies are used to organize content on various topics, though the details of specific implementations can be complex.  In an effort to distill some of the essence of what ontology is about, I read a recent book by an analytic philosopher named Nikk Effingham entitled An Introduction to Ontology (Polity Press, 2013).  I learned that ontology is a controversial field, full of debates and disagreements about terminology.  While the details of the philosophy were less practical than I might have hoped, I did find the range of topics debated useful to understanding many foundational issues content strategy professionals must take into consideration.

To get a sense of the importance of ontology, consider the emerging field of the Internet of Things.  Already, we have a multitude of things that are connected together, each sending out signals indicating what each thing is sensing. The devices and the systems they interact with, are constantly sensing, monitoring, and interpreting.  These signals may address social, physical, biological, or cognitive phenomenon.  What do all these signals mean?  Deciphering and translating their meaning is partly the role of ontology.

There are plenty of complex definitions of ontology, but I will offer a more direct one.  Ontology is simply identifying and describing what exists that might be significant.  We can refer to what exists generically as a thing.  We need to define what is the thing is, especially since we commonly define things in terms of other things.

When I looked at some of the philosophy that inspired this way of thinking, I found several themes that resonated with me that seemed of practical value.  I will summarize these below in several propositions.  They are my insights, rather than a summary of the field, which as I noted, it far from consensus.

Before we try to construct an elaborate network of inter-related definitions, the formal mechanics of ontology, it is important to identify dimensions that can impact how we define things.

image via synsemia
image via synsemia

Use Properties To Help You Establish How Similar Things Are

We can identify various properties associated with a thing.  The more properties a thing shares with another thing, the more likely the two things are related, all things equal.  Related things will be seen as complementary or competitive with each other.

The properties should be meaningful and hold significance.  Trivial properties (for example, something true in all cases) won’t convey much useful information.  The property that is most unique about a thing is often interesting, although it is also possible the property is not significant.  A test of significance is looking at the potential explanatory value of a property.  Might the property influence the behavior of the thing, or perhaps other things that interact with it?   Suppose we divide things into those that have metallic finish colors, and those with matte finish colors.  Is that difference significant, or inconsequential?  The significance of properties will depend on the context.  Rarely does one property alone entirely sway outcomes, so the covariation of properties are most interesting.

Differentiate Whether Things are Equivalent, or the Same

Sometimes a single thing is described in different ways.  Sometimes different things are described in the same way.

We know a single individual can have different monikers.  Mark Twain was the same person as Samuel Clemens, but unless we are aware that this author used a pen name, we might believe the names referred to different people.

A more challenging issue is when we want to see things that are broadly similar as being a single item.  We can match up things that agree on numerous properties and will say they are the same kind of thing.  When every property is identical, we can be tempted to assume the two things themselves are identical.  But we can be tricked into seeing things as identical when they are merely equivalent, due to the limitations of our descriptions.  Suppose all Model Z computers share the same properties we have identified — they have the same specs.  There is a bug in the computer, and the engineering team develops a fix.  The customer service team announces that Model Z’s problems are now fixed.  But a group of people using the Model Z continues to experience problems.  It turns out their computers are not exactly identical to other users: perhaps they have loaded software from Adobe or Oracle, which is outside of the specs the manufacturer tracks, which created the conflict.

Any time there are two or more instances of a thing, there will be some variation.  Sometimes that variation is so minor we can comfortably say the instances are effectively identical.  It is useful to know how much variation might be possible before assuming all instances of a thing will behave the same way.  And it is very important not to treat clusters of things (including people) that seem broadly similar as being identical.  There’s much value knowing how things might be equivalent to one another, but one should also be aware of potential differences.

Distinguish Categorical and Qualified Descriptions

Many descriptions are factual, and not subject to interpretation or subsequent change.  Your car will generally have the same engine size over its lifespan.  We say your car is a V-6: it is part of the identity of the car.  For many tangible things, as long as the thing exists, its properties will remain the same, even when it reaches the landfill.

Some descriptions are qualified by time or place.  When we see a map on a sign saying “you are here” we know that  “here” is relative to our current location and changes as we move around.  If we were to view this sign through a webcam remotely, the message would be incongruous.  As mobile technology allows us to shift location and time zones and communicate asynchronously, descriptions of where and when become more challenging.

Time and place can have more subtle effects on identity.  I once worked with a telecommunications firm that had a “family” package.  The marketing staff liked how friendly the word family sounded.  But when defining family, the definition subtly shifted to becoming members of your household.  Then the question arose of who qualifies as a member of a household.  Would children in university count?  If so, would it be only when they are living at home, or when they are away at university?  It may have seemed like an arcane issue to debate that distracted from other tasks, but the issue had significant impacts on sales, recurring revenues, and cost of service.

Ontologists refer to qualified descriptions as indices.  With the rise of big data, we are finding more indices pretending to be solid things.  I shop online, and am told that based on my “profile” I presented with various recommended products.  If I regularly shop for a wide variety of products, my profile is always changing, but never seems to match me, because now I’m seeking something different.

Know Whether a Concept is Abstract or Transitional

We often use intangible concepts to describe things — it helps us make sense of the qualities of a thing.  The meaning of many concepts we use to describe things are stable and familiar, so much so we think of them as real things, rather than as concepts.  When we say something is a meter long, the meaning of a meter is won’t fluctuate — the definition of a meter has been fixed for a couple of centuries.  Such concepts are abstract: independent of time or location.   But some concepts are less fixed, and more subject to time and place.  We might describe a work of art as contemporary, because it was created less than 20 years ago.  But in another five years, it might be more appropriate to call the same unchanged item of art as being modern, especially if the artist were to die.

Cultural values are especially susceptible to changes in meaning over time — just look at how old advertisements describe products in ways we find offensive or clueless today.  A simple example would be the shifting meaning of the term “healthy.”  Occasionally cultural changes can happen even faster than products change, such as gender role changes: for example, eyeliner for men being dubbed guyliner. Even many technology descriptions are conceptual, and transitional.  The label smartphone is just a concept that has no stable identity.  What we consider to be a smartphone has changed over time, and there is no guarantee we will continue using this term in the future.

Concepts are useful.  Just be aware that because they are often not precisely defined, their meaning is more likely to drift over time.

Watch Out for Frankenobjects

People in the marketing world are sometimes prone to package together unrelated items.  Consider Amazon Prime.  What is it, exactly?  Is it a club membership?  A prepaid shipping fee?  A streaming video service?  A music service?  What will it be next year?

Philosophers studying ontology call things that are glued together from parts that are not conceptually related or normally connected as “gerrymandered objects.”  Like gerrymandering in politics, the motivation is to trap something in an incomprehensible identity that changes form over time.  If you have to present a gerrymandered object to audiences, be prepared to do a lot of explaining what it is, how it works, and why it matters.

Understand What a Grouping has in Common

There are two types of groupings: collections and sets.

Collections are groupings of things that have common properties.  They are things that seem to belong together.  We see a collection of clothing for spring that are colored yellow.  The types of clothes are different, but they are all yellow, and all for spring, so they seem like a meaningful collection.  The creation of collections relies on rules based on the properties of things included in the collection.

Sets are things that are placed together on the basis of a choice that may be extrinsic to the properties of the items.  A good example is an online shopping cart. I may have a roll of cellophane tape, a pair of socks, and a bottle of allergy medicine in my cart.  It is a set of things that are unrelated, except for the fact I placed them there at a specific time with the intent to purchase them.  Sets are often created or changed as a consequence of an event.  There may not be any rules about what can be included in a set.

Sets may include related things, but do not have to.  Sets require interpretation to know what they contain, since we may not know the details or themes of their contents except through inspection.  We can combine sets, or look for unique items among several sets.  Whereas common properties define collections, with sets, you might check common or unique properties after making changes to the set.

Both collections and sets are useful, but they serve different purposes.

Understand the Changes of State that are Possible for a Thing

It can be difficult to say when something changes from one state to another.  This is especially true if we can’t identify a specific event responsible for causing the change.  I don’t know when my hair turned from brown to grey.  In fact, when listening to other people’s opinions on this topic, there seems to be minor disagreement.  Not only is change sometimes hard to pin down, it can be subjective as well.  When will my hair stop being grey?  When I shave my head, or dye my hair orange.

We assume hair color will change over time, even though it is generally not a significant property of people.  But we often underestimate the changes in the more significant properties of other things.  This poses two issues. One is that the description fails to update itself when the thing has changed. Another is that we are unprepared for unexpected change, and don’t even have vocabulary ready to describe it.  We need to account for edge cases, and intermediate properties that could be significant.

Closing Thoughts

Creating an ontology is challenging work. They typically require a team of people working over the course of years to develop them.  No one is going to create a new ontology for the Internet of Things in a few weeks time.

But ontological thinking is easier to do, and more immediately applicable.  Ontology reminds us that we often bring a point of view that colors how we perceive and categorize things.  Our view may be influenced by a specific time, place or situation in which we are located.  When we are aware of these factors, we can develop a person-independent description of things.

— Michael Andrews