Identification Pattern
This is work in progress, and needs to be reviewed. If you want to comment, please raise an issue.
Purpose
The Identification pattern captures the concepts and relations that help to explain the mechanisms that a party uses to identify entities, and mechanisms for communicating with another party such that both parties can identify an entity and know whether or not they identify the same entity.
Related topics can be found in the identity pattern and the identifier pattern.
Introduction
In 'the wild', terms such as identify, identification, identity are defined such that the can be used without all too many problems in everyday discussions. However, if we want to use these terms for our purposes, their definition must be much more precise, particularly because we want to use them in IT-contexts. Lets show this by example.
Here, we discuss identification in a way that we expect to be useful in IT-contexts (or non-IT, but still administrative contexts). Basically, two things are needed:
First, identification of an entity is similar to recognizing that entity. A bit more formal: given an entity (or digital or non-digital data about that entity that comes from outside the knowledge of a party), identification is determining whether or not the party owns a partial identity of which that entity is the subject. When the entity isn't recognized (identified), then a partial identity may be created that has this entity as its subject.
Second, identification of an entity is similar to pointing/singling out the entity. More formal: given a partial identity that is owned by a party and a specific context where a set of entities are present, identification is determining whether or not the subject of that partial identity is in that set, and if so, which of them it is.
We will elaborate on both.
1. Identification: selecting a partial identity
Suppose you are at home and the doorbell rings. When you open the door you see a person of which you will gather further characteristics (properties), e.g. by observing their looks, by characterizing that person's voice (pitch, intonation), and interpreting what the person says, etc. You may even elicit further observations, e.g. by asking things. All of this serves the purpose of 'establishing the identity of' (Merriam-Webster, identify, 1a) that person, where 'identity' would be 'the distinguishing character(istics) or personality of an individual' (Merriam-Webster, identity, 1a).
You could also phrase this as recognizing the person, i.e. tapping into your own knowledge to see what you already know about this person, which - in terms of our identity pattern would be: determining which of the partial identities you owns have the person at the door as its subject. The purpose of all this is that as you remember what you know about the person, that helps you to decide whether or not, and if so, how to further engage with him/her. This would mean that you gather characteristics of that person (the fact that it actually is a person being the first of such characteristics), and continue to do so until there is only one partial identity that has all these characteristics.
Continuing the example, it could be that you have no recollection of that person, i.e. there is no partial identity in your knowledge that has all these characteristics. In this case you would already have started to create a new partial identity for which that person is the subject, and you would have started to attribute characteristics (with some level of certainty) to this person, which all are part of this new partial identity. Doing this enables you to identify/recognize this person the next time (s)he comes to your house.
The identification process is said to fail if a user record or partial identity has been selected, but its actual subject is not the entity that supposedly is identified. This is the case when someone/something (successfully) masquerades as, or impersonates that entity. The process of obtaining assurances that the entity that is identified is "who (s)he says that (s)he is" is commonly called 'authentication'. We would say that authentication is the process of providing a set of assurances such that the risk of having selected a partial identity or user record whose subject is not the entity that was identified, is acceptable to the owner of that partial identity or user record.
2. Identification: selecting an entity from a set
Suppose you want to have a drink with your friend in a cafe. When you open the door of the cafe, you visually and auditorily 'examine' the set of people that are present, perhaps calling out to your friend, until you recognize one of these people as your friend. Having 'identified' your friend enables you to interact with him/her as you planned.
A bit more formally, we would say that what you actually do is acquire data (observations about the various people that are present), compare this with what you know about your friend (i.e. the partial identity about your friend that you owns), and when you find a match, you will have identified your friend. Note that this is the converse of what we described in the previous subsection.
Similar to before, this identification process is (also) said to fail if an entity is selected that is not the subject of the partial identity that you were looking for. And again, to reduce the risk of such failure to an acceptable level (for the identifying party), assurances may be obtained as necessary (authentication).
Identification in IT and/or administrative contexts
In the above examples, we tacitly assumed that the party is also an actor (see that party-actor-action pattern for details), which implies that the actor has access to all knowledge of the party, which includes all partial identities.
In many contexts, this assumption does not hold. Examples include IT-contexts, or organizational administrative contexts, the difference being that the agent that is doing the identification is not the same entity as the party that owns the partial identities that are being used in the process. In IT-contexts, an agent could be a web server, or a mobile app. In administrative contexts of an organization (which could also be IT-contexts with web server and other agents), you could also encounter human agents that are tasked with identification.
In any such cases, an agent that performs identification, must have access to the relevant subset of its principal's knowledge, that is: to a representation thereof that it is able to interpret as intended by its principal.
Such representations can come in many forms. In a human administrative context, such representations can be stored in e.g. customer and/or supplier relation management systems, employee registrations, etc., which may be paper-based (in a file cabinet), or electronic (in a database). Typically for IT-contexts, this would also include user registrations, that hold accounts for the users. The subset of attributes that one can find a such a registration should have been chosen such that the various records therein are fit for purpose, which means that any action that an agent of the owner of that data must execute, can find the data it needs for that task in one of its registrations.
This is particularly true for the action that is commonly referred to as 'logging in'. The actor that is 'logging in' some user onto a system (or provide the user access to some other physical or electronic location), will identity (and subsequently authenticate) the user, i.e. search for the account to associate the user with, from where it can find all other data that is needed about the user in that particular context.
Formalized model - information realm
Here is a visual representation of this pattern, using the following notations and conventions:
The figure shows that a party can know (about the existence of) any number of entities, one of which is that party itself. As explained in the party-actor-action pattern, this party has (owns, governs) its own knowledge. Also, as explained in the identity pattern, partial identities are a part of that knowledge.
As mentioned above, identification is an act, by (an agent of) a party that results in the establishment of a 1-1 relation between an entity (that exists in the real world) and a partial identity (that exists in the knowledge of that party). This act takes place in a specific context/situation, in which a set of entities is present.
Then, this act accumulates a set of characteristics (the 'uses' relation in the figure) that it either
- observes from the entity to be identified, then finds all partial identities that have all of these characteristics, or
- takes from the partial identity whose subject needs to be identified, then finds all entities that have all of these characteristics
There are quite a few commonly used characteristics. One example are entity-types, i.e. classifications of entities that parties use to make sense of the world. Examples of which include 'person', 'organization', 'animal', 'friend', etc. The partial identity that a party owns about some entity would include the entity-types that this party has attributed to that entity. Another example are identifiers, i.e. texts or words that are used to refer to an entity in specific contexts.
This process of accumulation of characteristics (and selecting the entities or partial identities that keep fitting all of these characteristics) continues until
- there is no entity or no partial identity that matches all accumulated characteristics, in which identification failed.
- there is a single partial identity and a single entity that both match all such characteristics, in which case identification is successful and the subject-relation between the entity and partial identity has been established.
Formalized model - data realm
Here is a visual representation of a similar pattern, using the following notations and conventions:
This section needs a figure and an explanation of how identification works in the data realm, i.e. the realm where identification is performed by an actor that is not the party on whose behalf the identification is being made. This requires that the actor can access data-records (owned by its principal) that represent (excerpts of) partial identities of the entities to be identified. Such data-records would contain attributes that represent the characteristics.