Introducing aggregates

The notion of aggregates is key to what I have to say in the next few posts.

Introducing aggregates

Consider two types of web page:

• Aggregates are lists of links: pages whose purpose is to provide a selection of options for further reading or viewing. Examples include portals, search results, and subject indexes in news sites

• Linear content pages, on the other hand, are essentially self-contained: they provide a unit of experience that can be consumed from beginning to end without going elsewhere. Examples include articles in news and magazine sites, video clips and some advertisements.

This distinction (inspired by Lev Manovich's between the “database” and “narrative” forms) should be taken as one between ideal types only. Of course few pages belong strictly to one or the other type; portals have more than just links, and good blogs provide a mixture of links and narrative (in fact, a characteristic of good web pages is that they mix these two types seamlessly). But, to a certain extent, every page on the web lies somewhere between these two extremes (the possible exception is  ‘search’, on which more later).

Note that, so far, all that is needed for these two constructs is a (a) system that allows for creating pages with stable addresses that can be referred to, and (b) a means of linking between pages. The web of course provides these mechanisms, but so do other new-media such us CD-ROMs, IPTV and even Teletext.

Aggregates and user behavior

The user experience involved in consuming pages of one type or the other differs in several respects:

• Prior knowledge: Consuming linear content presupposes knowledge, perhaps only of a general sort, of the subject, quality and style of the page to be read; and a decision to consume the content, or at least to try it. Consuming aggregate content, on the other hand, involves a less ‘intentional’ attitude: we don’t know what we will find on a search results page or a news portal. Part of aggregates’ value lies precisely in their unpredictable nature: we rely on them to tell us what is out there.

• Precedence: In the user experience, aggregates precede in time the consumption of linear content. We arrive at linear pages through links that mention them, and these are usually found in aggregates. Aggregates thus provide the context to linear content.

• Dynamism: Aggregates are usually changing, while linear-content pages are essentially static. A list of headlines (e.g. on a news portal) changes on a daily basis, but news reports rarely do.

• Recurrence: Because of this, aggregates are visited again and again, while linear content is rarely visited more than once. There is usually little point in reading a news report or a magazine article twice.

• Longevity: Because of this, linear pages' popularity exhibits a long tail over time: most of their visits occur in a relatively short time (the spread of the curve depends on genre; for news most hits occur within 3 days). On the other hand aggregates, especially always-changing ones like portals, have a more or less constant hit-rate. (Of course this need not always be so; long tails can always have an unexpected spike, and portals can gain or loose popularity with changing quality and marketing)

Aggregates, thus, are destinations: remembered landmarks that are returned to through memorable URLs or bookmarks. There is usually little point in having a short URL for a news report (although this does not apply to all linear content--for example, songs are revisited again and again over many years).

This characterisation of aggregates should also be taken as an extended definition. There are certain kinds of linear pages, such as some blog posts, whose main purpose is to provide lists of links. Although we could say that these are cases of aggregates, I prefer to restrict the term to pages that satisfy the criteria above: intentional, memorable destinations that people return to time and again precisely because they always change and link to linear content.

The values of aggregates

Users get several values from aggregates. At the most basic level, aggregates provide recommendations: in an age of infinite choice but limited time, we need reliable referrers to tell us what to read (as someone said, infinite choice means no choice). But to leave things here is to miss the point.

Aggregates are also community enablers: by keeping abreast of what is mentioned in them, their audiences not only get to consume content they appreciate; crucially, they also keep abreast of what the rest of the audience is reading, viewing or watching. Aggregates provide shared experiences. and content becomes a talking point, even if for quieter readers it is only an imaginary one. (The communities in question are of the type that Daniel Boorstin called "consumption communities")

Further, references in aggregates are always made in a crediting or discrediting light, however implicit. Because of this, and because aggregates always precede linear content in the user experience, aggregates crucially influence the pre-understanding that informs the linear-content experience. We experience content differently depending on how we arrive at it. An important consequence of this is that content abstracted from aggregates is worthless. This post's fate, for example, depends crucially on who mentions it and in what light the referring is done. (This also applies to old media, a fact that music marketers have been keenly aware of for some time—think of their relationship with radio DJs).

The meanings of aggregates

All of this can be seen in a more coherent light if we think of aggregation as a creative act, and of aggregates as a form of content with its own idiom. Consider then what is implied in the act of including an item in an aggregate:

First, inclusion implies a declaration of relevance. By including an item in a list, I am implying that it is relevant to my readers and that I recommend visiting it. This in itself says nothing about the quality I attribute to the content; I might be referring to a bad song or an uninspired piece, perhaps mockingly or to illustrate a point. My promise to my readers is not that the content is ‘good’ in its own right, but only that its consumption will be coherent with my aggregate.

Second, the way in which I refer to content may reveal some of my appraisal of it, an appraisal that I invite my readers to share. This may be of various kinds (mocking, praising, despising, agreeing or disagreeing, etc) but will generally either positive or negative.

Finally, most aggregates (RSS feeds are a notable exception) also allow for a sense of prominence—i.e. a way of saying that some things are more important than others. At its most basic, this takes the form of ranking: some items are displayed before others; in other cases, prominence can be expressed through larger space (physical or temporal), various graphic devices, or an explicit declaration of importance.

Aggregates are thus a form of expression, a form with a richness comparable to that of linear content, i.e. to that of narrative. They are a distinct authorial form that audiences value in its own right.

The navigation orthodoxy: search and other automatic aggregates

There is a school of thought that is likely to dismiss the claims made so far. According to this view, aggregates are just navigation tools, that is, ways of allowing users to get to content; they are not forms of expression. Authorship in aggregates is at worst obstructive (bias), and at best useful as a recommendation device.

There are two problems with this view. The first is that, while it is correct, it accounts for only a fraction of media experiences. The second is that even in these cases authorship rears its head.

• The image of users ‘getting to’ content can be broken into two scenarios. In the first we have users as seekers—e.g. of an article or song whose title they can't remember. In the second we have users as profiles: people with certain characteristics (e.g. demographic, behavioral), and perhaps some vague desire, who need to be matched up with appropriate content. It is undeniable that both of these scenarios occur all the time, and there exists tools to suit them. The first is best served by search; for the second, tools like directories and recommendation engines do the job.
  
  But to claim that these scenarios account for all or even most user experiences is to leave out a large part of media consumption. The first scenario presupposes interest in a piece of content; this interest in all likelihood originates either in an aggregate consumed earlier (perhaps in a different medium, such as a magazine), or in a word-of-mouth chain that at some point started in aggregates (although this is sometimes a crude simplification; to say that I heard of King Lear through word-of-mouth is to reduce culture to marketing). The second scenario suggests content that will be desired even in the absence of prior knowledge and recommendations—a situation that might be conceded to apply to some experiences (e.g. someone in search of entertainment, or a researcher looking for information), but certainly not to all (e.g. the relaxed reading of a newspaper on a Sunday morning).

• Even if these concessions are made, authorship in aggregates cannot be altogether dismissed. A search results page may be generated by a machine, but the best algorithms (e.g. Google’s PageRank) rely on heavily mining authored aggregates to extract relevance, and are additionally tuned by teams of editors with a sense of relevance and appropriateness. Similar comments apply to recommendation engines; here too the best algorithms (e.g. collaborative filtering) mine authored aggregates (some of which may never have been meant for public consumption, such as purchase histories).
  
  Thus, while it is true that lists of links provided by search of recommendation engines are not strictly authored constructs, they are heavily informed by human-authored aggregates. This is not incidental; relevance is an essentially human, value-laden construct that machines may mine but never reproduce.

Finally, these ‘automatic’ aggregates provide values beyond that of mere recommendation; just like authored aggregates, they are community enablers. When I enter "the long tail" in Google I expect to find that article that everybody is talking about, not another article to which the phrase is more appropriate. Similarly, a personalized list of book recommendations is interesting not just because I may ‘like’ the books it mentions (like a cat would like a certain type of food), but also because I know (because I am told) that “people who bought this also bought that”, which may bring me closer to the community of interest that reads my book. (Incidentally, my cat does seem to mind how I present his food to him. If I make no fanfare then he's generally not interested).

All of this can be summarised by saying that aggregates are content: they are distinct editorial artifacts that audiences seek in their own right, not just as means to other content; they are creative products with their own value, style and personality. (For a clear example, see the "news wire" in The Hufflington Post).