I've
been waiting for Google, which has already provided a definitive
'know-what' information finder and 'know-where' place finder, to follow
up with the definitive 'know-who' people-finder. My initial thought was
that only Google and one or two other giants could get enough profile
with this to get everyone to participate and accept it as the standard,
and hence achieve the critical mass to succeed where so many Social
Networking tools that have tried to do this have failed.
But
then it occurred to me that there is a profound difference between
'know-what' and 'know-where' on the one hand, and 'know-who' on the
other: Finding the former are complicated search problems; finding the latter is a complex problem. Google can write an algorithm to point you to the documents most likely to be useful to you on subject x, and they can create maps to point you to location y. You don't have to do anything but ask. And although the numbers are vast, there are only a finite number of documents and places on the planet.
By
contrast, any meaningful people-finder would require active and regular
participation of many people. Assessments of expertise are too
subjective and change too quickly over time for any kind of algorithm
to 'interpret' from data that are already 'out there'.
But any
kind of top-down, managed expertise-finder will inevitably fall victim
to the same problems that have afflicted Linked-In and other social
networking apps: Not broad enough participation, data that is stale and
which no one is motivated to keep current, and the tendency of people
to try to 'game' the system to portray themselves as more popular and
expert than they really are.
The only thing that will work, I
believe, is a Peer-to-Peer solution, one that works with existing
ubiquitous tools and which makes it easy for anyone, regardless of what
platform they are working on, to participate with little or no
incremental effort. When addressing any complex problem, we need to
give the solution the opportunity to evolve as the result of the
collective intelligence of everyone.
That is an imposing
challenge but not an impossible one. What we need to do first is
develop a high-level spec for a system that no one
will build. The spec will be merely the initial set of principles and
guidelines that will influence how we participate. The 'crowd' will
tell us if some of those principles and guidelines are wrong, and
what's missing, and we'll change them to reflect that wisdom and
imagination.
Here's my first cut at some of those principles and
guidelines. We need the people who know the Internet best, both as a
technical and social phenomenon, to add to this list -- we won't get it
'right' the first time, but the closer we get the list in the early
stages (or, to use complexity terminology, the more valuable our
initial set of attractors and barriers), the faster something useful
will start to emerge from it.
- The terms for expertise should be folksonomic, not taxonomic:
It is futile to try to design a taxonomy of expertise -- there are too
many terms and types and they change too fast. Let everyone decide on
their own terms to define their, and others', expertise, and let the
system accommodate them. This is like what last.fm does with its music
tags -- it doesn't set out a predefined set of genres, it allows its
millions of users to self-define tags that mean something to them, and lets the 'crowd' settle the matter. A critical corollary to this principle is that the terms are not hierarchical:
Your terms for expertise can be as general or as specific as you want.
Some people may be looking for generalists, others for highly
specialized people -- the system doesn't discriminate.
- People can define their own expertise:
The best initial set of terms of expertise is probably self-defined: We
all know ourselves (or think we do) better than we know others. We
don't get to 'vote' on our own expertise, but the best way to initially
populate the folksonomy of expertise terms is to get a few million
people to 'tell us about yourself', to define their personal genius in
their own terms.
- Voting on others' expertise should be simple:
Forget 1-10 scales. The best gauge of expertise, one that is
independent of the financial wealth of the voter, is how much time you
would be willing to spend listening to and learning from that other
person on that subject. Your personal time and attention is the
ultimate investment, and your willingness to invest time and attention
is hence the ultimate measure of another's expertise. So the vote for
others' expertise should be as simple as I would be willing to invest (a) lots of time [H], (b) a little time [M], (c) no time [L]
with this person on this subject. There is no 'default'. You don't vote
on another person's expertise unless and until you know them well
enough to intelligently answer the 'I would be willing to invest time'
question.
- Your votes would sit and be maintained on your own hard drive:
No submissions to central repositories -- we've seen again and again
that that mechanism just doesn't work. Last.fm gets its data by
automatically harvesting data from participants' iTunes as they play
songs on their PC -- no need to 'tell' the system what you like. We
need a place that everyone with a PC has in common, and my suggestion
would be the Address Book. Add
to, or requisition, a field from the address book to use as the
expertise assessment field. So if I think John Smith is a terrific
expert in cultural anthropology, an interesting guy to chat with on
innovation, but less than useless in his self-proclaimed area of
expertise, social networking, I'd enter those three folksonomy terms in
the expertise assessment field of his e-mail record on my hard drive
followed by an H, M, and L respectively. I don't know whether address
books do, or can be made to, date-stamp when these assessments are
made, reconfirmed and changed, but the date of this assessment is also
pertinent and needs to be captured somehow.
- Different expertise 'scores' are needed for different purposes:
Suppose I'm looking for an expert on 'knowledge management' in the
'health care' industry. I may want to know who has been rated 'H' by
the absolute highest number of people in both those areas, or rated 'H'
in a single tag 'health care knowledge management'. I may instead want
to know who has the highest median rating (2H+M)/(H+M+L) in these
areas. I may want to know either of these things but counting only
assessments made or reconfirmed in the last month. There is no single
'score' that meets all needs, so any scoring algorithm needs to
accommodate these different needs.
- We need to be able to filter and analyze expertise assessments in many ways:
We may only want to see experts who live in certain geographic areas or
who speak certain languages. We may only want to see experts who will
give us some time or expertise free of charge, or whose rate is less
than, say, $50/hour. And we may well want
to see the identities of the people making the assessments, and
discount those in large organizations who rate everyone else in their
organization 'H' on everything to 'game' the system. The system may in
fact evolve to allow us to assign a trust/credibility rating to
different assessors, and filter out assessments from those we don't
trust.
- We should leave the tabulations to those who do them best:
Rather than trying to come up with our own tabulation system, we should
simply charge Google, Yahoo, Technorati and other companies that are
already expert in search and ranking algorithms with the task. As the
tabulators of 'know-what' and 'know-where' information they have a
vested interest in tabulating 'know-who' information as well. They can
also grapple with the security issues (e.g. accessing people's Address
Books to harvest, or canvas just-in-time, the assessment data to
respond to 'know-who' search requests).
I can see this
evolving in interesting ways. Corporations will initially want to use
this within their Intranet firewalls to find experts within their own
organizations, and won't want
that data accessible outside the firewall. But information is always
trying to be free, and once smaller organizations 'let it out', and
buyers start looking for and expecting to see their preferred
suppliers' experts' names showing up on 'know-who' search results, the
big professional firms will have no choice but to open up the data to
the world and let buyers start putting together their own
cross-organizational teams of experts.
I also think that being
acknowledged as an expert is a double-edged sword, and such a system
will start to create genuine 'markets' for expertise. People
acknowledged as experts who are bombarded with requests for their
expertise, and who cannot afford (or do not want) to spend their whole
life sharing what they know free, will naturally start to put in
personal, market-driven hourly rates for their expertise, and hence filter out most of the requests.
Who
knows, some of us might find that we're acknowledged as experts by more
people than we think, and we might even be able to make a living simply
on the strength of this system's 'word of mouth'.
It sounds very
complex and unmanageable, I know, and it is, which is probably why it
hasn't happened already. But there is a clear need for a viable,
simple, reliable, easy-to-maintain expertise finder, and once a few
million people agree to start maintaining the information that would
drive it, I think it could explode quite quickly, and evolve just as
quickly to meet this need extremely well. The key is not to try to
design a centrally-managed app for it, but rather to let it grow and
become what it will become, virally and organically.
I'm going to pass my thoughts along to Doc and David W. for a start (since I recognize their
expertise in this area as 'H'). If you think this is a useful avenue
for exploration, please talk it up and tell me what you, and others you
talk to, think. |
MADE IN CANADA
trust your
instincts 
