Photo credit: Kirsty Pargeter
Creating feedback loops
One of ebay's and Amazon's best features is the "web of trust".
These interconnected webs, that simply report a summary of other buyers' feedback about a provider, are key to the new web. Where they don't go far enough is that they're limited only to the data on each discrete island of information - the individual site - because of problem #1, the disconnected web!
Other limitations of the current state-of-the-art are:
- The feedback systems are relatively crude.
- They give an overall good/bad rating, but can't positively identify the most suitable solution out of a good bunch.
- Not everything fits into auctions or e-commerce, or even one-off transactions. For example, I can't see my cleaner selling his or her services on ebay.
All we need to do to get a killer marketplace is:
- Create a common framework that lets anyone promote anything anywhere, and let other people search for whatever they want (Google Base, Google Classifieds and Microsoft Fremont are starting to look into this area, although unclassified ads will be more useful than classified ads)
- Record feedback whenever someone uses an offering (product or service)
- Use contextual information on suppliers and searchers
- Recording the results of matches (e.g. "You looked at Brigg's Builders. Rate your service...") will help the system build up a useful, rich picture that will help the system predict the best matches for future searches.
On one level, you get the crude "95% positive rating" metric. However, just by keeping and joining up a bit more information, we can do a whole lot more.
Remember, instead of trying to arrange our universe of information into an arbitrary single line of best-to-worst, the goal is to predict the best few matches to a particular need in context, i.e. at a particular point in the interconnected web in space and time.
Case study revisited
So now if I do a search for someone to help me clean my house, the killer marketplace system can easily see:
- You live here in the world.
- You're looking for a cleaner, which is a distance-dependent service, therefore locality is important (not hard to learn this stuff from people's searches).
- I know 30 people promoting services as cleaners in the local area.
- Five of these come generally well-recommended.
- You are connected to the Neale family (via affiliations Baslow School, and through the Bulldogs Soccer club). Therefore you're likely to have things in common with this family. They gave one of the suppliers a poor rating.
- You are also well connected to Fred Brigthman (by being IM buddies, living nearby and having used the same car mechanic and tool shop), who has given another supplier a very good recommendation.
- This will give us our answer - the one that Fred Brightman used and thought was good. It may not ultimately prove the very best answer, but it's the best that's possible with the information that can be known at the time.
The process I've noted above is not at all complicated, it's just simple analysis of raw data. However it lets us predict with a good degree of certainty which cleaner out of the 30 near me will be most likely to suit my needs.
Therefore, my search doesn't need to return 30 or 300 results. Because it's smarter, it can just give me the top 5 to try.
Again, there's no new technology here. It's just existing stuff, but joined up.
And it can work, easily! (I was playing with a working design for this interaction over a year ago, which solves the issues of trust.)
Let's consider how the system will learn this stuff about me in the first place. (We can't expect every commercial web site out there today spontaneously to share their data together!)
Future killer contacts manager
Again, there's nothing much new here. All our future killer contacts manager needs to do is collect some of the same information we do today, and let it move about more easily.
A few things on the market today point toward what's possible:
LinkedIn, Spoke and others are basically databases of people and the connections between people. They enable services such as finding contacts in particular areas or in particular companies, or connecting to new people with similar business interests to yours.
Plaxo lets you store your contact information in a single, central place. If your contacts also use Plaxo, then when you update your contact details, their copies of your info is updated automatically.
These are great and well-designed web products, however they still suffer from the general disconnection disease.
Here's a vision of a contacts manager that's transparent to use and does a lot more than simply record who you know.
The qualitative differences are:
- You can connect to more than people
- Connections can vary in importance
- Some connections are only valid when reciprocal<
This image represents a tiny area of the connected web. It shows a number of elements, which may be of any type, connected with simple links of varying strength (and each having its own extensible database in XML format).
How you connect to people
You'd create connections through a number of entry points, each with a similar result:
- By sharing emails
- By adding a messenger buddy
- By right-clicking/option-clicking a person's smart-tagged link on a web page (see below)
In this example, you could be reading a blog or post by someone. Their name is marked up with their ID account identifier, which lets your browser offer you the option to connect, with a "connection strength" that reflects the value you place on your connection with that person. Note that this doesn't create any connection from that person to you.
Here's how it might work in a desktop email package (Outlook in this example).
You can connect to more than people. The power of connections is key the connected web. The reason it's so powerful is that it represents the connections in real life. I cannot stress enough how important this point will be.
In real life, I can be connected to more than other human beings. I can also be affiliated with:
- Companies and brands - e.g. I may choose to express my connection to Apple computers, SitePoint books or my business coach Ali in terms of trust, support, and fidelity.
- Products and services - As we've already seen, if I rate a product or service, that's real data that can be useful in building a picture of individuals and the links between them.
- Groups and organisations - membership or support of any political parties, clubs, societies, religious groups etc. are also connections.
- Web sites - If I bookmark and rate a web site (as good or bad), I'm creating data, which could be useful knowledge.
All this information is real and, taken together, can be used to get a quick heat map of my place in the connected web.
In the connected web, all these entities can exist alongside people. You can connect to them, and they can connect to you (recognising you as a member, family member, affiliate, subscriber etc.).
Connections can vary in strength
Most existing systems simply collect links, saying "Ben is connected to Thomas. Thomas is connected to 2,000 other people. Therefore, Ben has a million potential friends."
This has limited use, because it doesn't reflect the quality of the relationships. In real life, the relationships and links that connect us to other entities can be weak or strong, and also positive or negative. Therefore, links on the connected web must also reflect strength.
Let's say I get an email from a designer working in China, who asks me some interesting things. Using my connected email client, I may think, "You know, I like the sound of this guy." and simply right-click > Link > 3/10. We're done! I have now expressed my link to Chen, my new contact in China. That link will be recorded on my central identity-account, where it can be used in all kinds of ways to benefit me.
This means that:
- I have recorded a low-level of recognition and synergy with another person
- His heat-map will affect my heat-map (a bit)
- My account can talk to his account so that I can always find his contact info (possibly), read his blog, see his homepage etc.
- Email from this guy always gets into my Inbox.
At a future point, I may change my mind about this contact, and may strengthen my link to him (if he grows on me), or set it to zero (if he turns out to be a weirdo). The important thing is that I have control.
Some connections are only valid when reciprocal
Let's say that my friend, Chen, wants to send me an instant message. I may set up my IM software to allow only people with whom I share a 20%+ link to make contact with me. This is fine, because I trust Chen 30%.
But some functions in social networking rely on mutual recognition, i.e. reciprocated links.
Let's say Chen has a friend, Danny, who's looking for a contact in Interaction Design. He may do a search on a connected web site, or through his IM client. This search will be effected through a peer-to-peer style interaction.
Here's how it works:
- His client contacts his ID account.
- His ID account contacts the ID accounts of all his contacts (i.e. the people he trusts). If those account recognize him (i.e. there's a reciprocal link), they may respond to his request. In this case, he's contacted Chen's account, which reveals that Chen is interested in Interaction Design. Because Chen trusts Danny 80% and Danny trusts Chen 60%, there is a 60% link between them (simply the lowest value). And because Chen allows his account to reveal general information about him to anyone with a total 10% link or more, the accounts talk, and Danny instantly sees a result: "Chen"
- Now, Chen's account passes on the request to its contacts.
- Chen's account contacts my account, because Chen recognizes me (50% link).
- My account recognizes Chen - with a link strength of 30% (the lowest link strength). Chen's account says "I have a contact, value to me 60%, who's looking for contacts in Interaction Design"
- My account calculates that the link from Chen's contact to me is 60% x 50% = 30% (a composite connection, reflecting the remote nature of my link with Danny).
- I permit my general contact info to be passed to people with a 20%+ path strength, so my account passes my general contact info to Danny's account, which passes it back to Danny's software client, and I appear as a result on Danny's search.
All the above would happen automatically in just a moment, and illustrates why some transactions in a social network rely on mutual recognition.
In the next and final part of this essay I will look at future killer security, spam management, other possible effective applications of the interconnected web like dating, ride sharing, groups and the commercial interests surrounding them.
First part here: How New Social, Interconnected Applications Give Way To The Untapped Potential Of The Semantic Web.)
Ben Hunt -
About the author:
Ben Hunt, aka the "Web Doctor", is a UK-based interaction design consultant who has been designing software, web sites, and web applications since 1996. Ben runs a small UK-based consulting business called Scratchmedia, and publishes 'Web design from Scratch' to share his passion for designing excellent user interfaces with a worldwide audience. His key mission is to help make the web a better place by learning and sharing the practice of good design.
Reference: Web Design from Scratch [ Read more ]