I've received some personal correspondence about the article I posted recently on ownership, structures and behaviours asking for a bit more detail on a few points so, rather than answer it privately, I thought I'd share my thoughts a bit further. My correspondent asks the following questions...

Could you give us some concrete examples of:

- deliberately parcellating the environment: how?

The theoretical basis for this comes from both natural ecologies and human systems such as city planning. In essence, larger flatter ecologies are less dynamic, less adaptable, more rigid and blander than smaller, largely separate ecologies; they offer little opportunity for speciation, novelty and variation so they tend towards stalinist regimes where little changes, making them brittle and susceptible to disaster if aspects of the landscape or surrounding systems change.

You get greater diversity through small, differentiated islands than through vast, featureless savannas. Creating greater parcellation increases the rate of evolution in the smaller spaces as species with different (and at first more limited) competition than that of the big flat spaces, different surrounding landscapes to adapt in, and different available resources to adapt to, fit the spaces.

As long as there are occasional or weak links with the larger set of ecological spaces (such as the odd isthmus forming), this in turn provides more possibilities for adaptation of the overall grand system as other things in the environment change. In the event of something bad happening in a parcellated environment, the effects may be contained. If something good happens, weak links and occasional ties allow it to spread. Good and bad are asymmetrical because really bad things have a tendency to wipe out or greatly reduce the activity of their parcellated niches, so they often don't spread as easily as good variations. It happens, of course - it just happens less often.

General examples of how to do this in a social environment include:

• provision for hierarchies (e.g. groups as members of other groups, taggable tags - I'm having a plugin for this built),
• adaptable interfaces (e.g. different themes for different areas of interest, configurable widgets and plugins for different group areas, the potential to present different façades for different contexts and audiences - another one on my list) and
• broader systemic interventions (e.g. using marketing to encourage take-up by different groups along with relatively isolated social spaces for them to inhabit).

There are concrete examples of how I have used this principle in the systems I have built - Dwellings being the one that most explicitly used it, but it was a principle that emerged through the research I did using CoFIND. Peter Brusilovsky's Knowledge Sea II uses a similar idea, and the general use of tags allows little ecosystems to evolve in the context of tags and tag groups - del.icio.us does this very well.

In a teaching context using available tools, for example, I used the customisation options of the old version of Elgg to create a visibly different space than the rest of the site for a class, with internal navigation explicitly provided, a different colour scheme, different tools and different menus. This was an attempt (not wholly successful) to overcome one of the main problems faced when attempting to build a group in a system that is largely centred around networks and collectives. These forms of the Many create their own structure and dynamics that are seldom in perfect accord with what we want to do in groups: it's way too easy to slip outside the group context and get lost in network-space, or to follow a collective aspect (such as a tag) and drift away from a focussed context. Peter Sloep and others at the OUNL have looked at similar issues in their work on ad-hoc learning networks.

Probably the most common way to parcellate is the most extreme - using separate systems and different sites for different purposes. For instance, you might handle your business connections on Linked-In, your personal ones on Facebook, and your educational ones on Ning. However, while that is great for getting Galapagos-like diversity, it also sets up far greater barriers to the spread of good ideas and adaptations between ecosystems. For instance, because of the interdependency between technology and behaviour on a social site, we will often have to reprogram a computer system if we want to make it mimic nice behaviours in another, or at least customise how it displays things, which is often hard and sometimes beyond our control altogether. Ignoring good ideas to limit the problem like RSS, Open Social, RESTful APIS and service-based architectures for a moment, the lack of capacity for good things to drift between islands makes this (at least until we are better at doing reliable, persistent interoperability and aggregation) a less satisfactory approach than building our systems with parcellation in mind.

- actively intervening: any examples?

For instance, by making use of the 'featured groups' facility in Elgg an administrator can actively guide users to particular growth areas, or de-emphasise an over-dominant use. We are doing this in the Landing to draw people's attention to meta-activities at the moment, to encourage active participation in the growth of the site. Active marketing outside the computer environment can also be used to encourage different uses. For instance, my concern about over-marketing academic uses could be compensated for by actively marketing (say) uses by societies and unions or setting up and actively cultivating groups to share jokes or complaints. Taking it to extremes, we might actively run a query on the database to change numbers applied in weightings and rankings - when building my early toy systems I used to do this experimentally to find out what happened when algorithms changed or, a couple of times on a live site,  to compensate for bugs or features in my algorithms that led to Stalinist or red-queen effects.

- showing random novel items: how, through what mechanisms?

This is purely programmatic and means changes to the ranking algorithms - rather than showing the latest/most popular/most active, to introduce some randomness in the algorithm and push random novel items. In one of my early systems (CoFIND) I used to give a specific (and dynamically decaying) novelty rating that took into account other activity on the system to be used in ranking algorithms.

- intentionally pushing other uses: any concrete examples?

See my comments on active intervention. An obvious example might be taking out an advertisment on another site to promote a new film club inside the site. We are putting out a call for participation through an internal magazine site in which we are deliberately offering an assortment of use-cases so that people do not simply see the thing as another teaching tool. When I have used Elgg groups in teaching, I've often nudged things in a different direction simply by providing a new example or particularly obvious new resource (my students have been known to hate me for the Blog Song, an annoying jingle inciting them to blog). If you can get a group of people to work concertedly on this kind of thing you can easily overcome most collective and network applications. The classic example of this is Google Bombing, but the same idea can be applied to almost any site that uses the network or collective as part of its dynamic design.

- making it easier to move to different spaces: how? Examples?

This comes from Jane Jacobs's ideas on what makes for a thriving city area. In the Death and Life of Great American Cities she identified short blocks as a vital component, because it makes it easy for people to get there from different places. On the Web, the equivalent is appropriate linking. So, for instance, you might add toolbar or sidebar links to areas you want to emphasise, perhaps as part of a strategy of active intervention where you want to promote a particular set of groups or behaviours, or perhaps because that's where people seem to want to congregate. Our developer at AU has come up with a nice innovation that makes the tools on the Elgg system more visible by putting them in a side menu, thereby making it easier for inhabitants of the site to get to the things they are likely to need most.

More generally, in interaction design terms,  it is important to make the passage from one well-populated area to another as easy as possible. I haven't found an easy way to use the crowd to automate this, though the 'what's popular' links on some systems are an approximation, albeit one that carries risks of mob stupidity setting in (see my notes on the importance of delay below).

- changing the relative weightings of novelty and popularity: How, with what tools?

That's another bit of pure programming. For example, the current tagcloud plugin for Elgg is a little basic, using nothing more than overall usage to give the weighted list. You could do a Flickr-like thing and, in addition, show the most popular recent tags as well, for example, or you could change the weighting of tags by adding in a boost for novelty (or taking away some weighting from tags that have been little-used recently). The same principle can apply to any dynamic items that are displayed on the site. I've built a couple of toy systems that allow you to control this for tags so that you can make up your own mind whether you look at the reflection of the overall group mind or concentrate more on what it is thinking about now, or select something in between.

- introducing new tools: examples?

This is about cultivating diversity. The more primary uses (reasons people have to be there) the more the system will get used. So, for example, by adding task management tools and calendars, you can encourage groups to handle some project management as well as sharing and communication. Or you could add a plugin to enable people to buy and sell used items on the site. We are trying to do these things now.

- increasing personal control: how? For what specific purposes?

This is a big one. Apart from the fact that control is a good thing in itself (it makes people happy to be in control), it is also a central component of ownership, and vital in cultivating diversity. To take it to an absurd extreme, if people had no control over their social environments they may as well not be there at all as they will have no effect on its development and structure. The more that people can control, the more they are likely to use a system and the more able they are to contribute to its growth. 

Control is, of course, not identical to choice - it's about allowing people to delegate the choices they don't want to make, as well as to enable them to make choices that they do want to make. Too many choices can take away control.

Bearing that in mind, we can allow control of many things including relationships (e.g. what we reveal, how we reveal it, how we communicate, how we connect, how we form and join groups etc), content (e.g. what we reveal and how we reveal it, what we use, what we don't use, how easy it is to discover new things), our personal learning environments (what tools, people, content, feeds etc that we bring together, and how), time (when we reveal things, when we choose to communicate), effort (how much energy we are willing to put into something) and much much more.

There are myriad ways to do this: things like Elgg's permission system, its drag and drop widgets, its support for RSS feeds, ease of participation through universal comments, and lots more besides make it pretty good in this respect already. We are working on increasing the number of choices (e.g. over which blog posts you reveal, not just whether you reveal them, or over the visual style of what you use and show, and to whom you show different façades) without reducing control - things will work more or less as they do now but we will just make it easy to drill down and change more things for those that want it. Our developer is also tweaking the interface to try to provide the things people need when they need them - for instance, when reading a Wire post, it is logical to make it as easy as possible to write one. Another nice small interface change he has made is to make it easy to spot the arrow that gives a drop-down menu that can pop-up whenever viewing someone's profile image in Elgg. I have also commissioned a tool to allow people to present different façades to different groups of people.

Almost every improvement we are trying to make is about increasing personal control, in fact. The only notable exceptions apart from internal changes to make things work better are cases where too much control by one person impinges on the liberty of another. So, for instance, we don't wish to allow total anonymity because that will reduce the trustworthiness and safety of the system for many people.

- adding delay: what do you mean exactly by this?

This one takes some programming, for the most part, and there's not much out there yet that takes advantage of the idea.

Mobs are stupid - people tend to follow the crowd. Partly that's a logical necessity (reduced choice caused by previous choices) and partly a psychological oddity of human beings. Things that get there first tend to have far greater influence than things that arrive afterwards, and there are positive feedback loops in any crowd-based system that mean that, instead of an aggregate wisdom arising out of multiple intelligences thinking simultaneously, the first moderate or fair idea that takes hold tends to remain that way even if it is low down on the overall fitness landscape.

Avoiding this problem is one reason for parcellation - it limits the effects of mob stupidity to a particular niche, and allows diversity to develop and alternative solutions to present themselves. Another way of avoiding mob stupidity is to introduce delay: to allow individuals to make decisions relatively unencumbered by the influence of the crowd. Once you have sufficient delay (the precise amount varies as a function of overall activity) then there is time for

1. a good consensus to form and
2. for alternative niches to develop. 

The positive feedback loops can then work in our favour, driving good things up and bad things down, achieving a bit of fine tuning and deeper order to the initial consensus. This is one of the reasons Google is not too bad at all: there is an inherent delay between linking to something and it being sucked into the PageRank algorithm. If Google had instantaneous access to every change on every website, and if it used it the moment it happened, the results would be far worse than they are, because the stigmergic effects caused by Google being the number one search engine would mean the first would always have a huge advantage over the next, which would be more likely to get a high ranking than the next, and so on down the list. As it is, there is time for intelligent, discerning individual decisions to be made so that, when things are sucked up in PageRank, they are (at least often enough to be noticeable) chosen because people like them as things in themselves, not so much the result of following a crowd. Note that we are talking about statistical influences here - not every individual acts the same way in response to the same stimuli (thankfully) but, when you look at the whole, there are definite tendencies that, in a crowd driven system, get magnified or amplified out of proportion.

This also relates back to my earlier comment about getting the right balance between overall popularity and novelty. If you check out Flickr's most popular tags in the last 24 hours vs the most popular overall you probably find the latest are almost unusable - it's a bizarre selection of rapidly changing oddness, well over the edge of chaos, practically unusable. If that's all Flickr gave you, tags would wind up being utterly useless. On the other hand, the overall most popular tags are 80% the same as the most popular tags 5 years ago and are frankly boring and not much use at all as each one provides you with hundreds of thousands or millions of undifferentiated tagged resources - it's too far into the other, orderly side of the edge of chaos. The sweet spot is just on the orderly side of the edge of chaos, where you get evolution and development but with sufficient stability (in this case caused by delay) to find out the things that work.