RE: The Object Primitive

I think the center of this conversation orbits the concepts that:
+ a value for the trusted nature of the data that comprises a Place and or a Location is valuable to the user of that data.
+ there will be many creative ideas on how to best compute, adjust and corroborate the value of trustiness.

At NAVTEQ we have the concept of a PlaceScore and a LocationScore that are a patented weighted trust/quality-of-data scoring based on regional algorithms for completeness, source trust level, geo-specificity, age, method of validation, etc.

_______________________________
Karl Seiler
Director Location Technology & Services
NAVTEQ - Chicago
(T)  +312-894-7231
(M) +312-375-5932
www.navteq.com

-----Original Message-----
From: Thomas Wrobel [mailto:darkflame@gmail.com] 
Sent: Tuesday, February 15, 2011 5:54 AM
To: Jens de Smit
Cc: Roy Davies; Seiler, Karl; cperey@perey.com; public-poiwg@w3.org
Subject: Re: The Object Primitive

I think it would possibly have to be done by some sort of independent site.
So, that is, people can subscribe to a site/stream that associates
"trustworthiness ratings" with other sources augmentations.
Extra server-calls for the user subscribed to the system, but I think
its more do-able then all the POIs being updated automatically, and it
leaves room for a few various types of trustworthiness evaluation.
This would, however, depend on the ability for augmentations to be
identified (and preferably other content to be positioned relative to
them.)

Regarding mana;
The system like that can work well in any Web2.0ish system. That is,
some successful current sites with
user submitted content that use a system of weighted votes, where the
weight of your vote is dependant on how many others agreed with your
vote.
You have to be very careful with a system like this...cap the votes
per day to a low number, monitor IPs to ensure there isn't likely to
be much self-voting going on etc.
Basically, it will be a war against spammers.  But its possible;
newgrounds.com pulls it off fairly well. A good 90% of the flash
animations submitted to them is either spam, or god-awful stuff, yet
they manage to get it filtered by a almost totaly user-voting based
system.  So its certainly possible.

I always thought weighted voting would work well for a semantic
database too, where the associations between items and property's gets
stronger the more people agree with it, allowing essentially a large
user-submitted "mesh" of relationships. It would be an interesting and
usefull tool, and I'm surprised no one has attempted it yet.

I'm getting slightly offtopic here though... ;)

Trusthworthy can also be a very individual thing too - where you value
the views of your friends and family more then strangers. Vark.com
works this way. (It prioritises your questions to friends and friends
of friends before it starts asking strangers).

Another system of Trustworthiness is googles "knol" - where academics
write article's of specific things with their name attached so you
know the source of the information specifically.

Yet another system for trustworthiness can also be based on the
"degree of agreement with others on other things". So, if someone
likes 5 movies that you do, its more likely that you would like a 6th
movie he does, rather then someone that hates those 5
movies....probably. This is how my own rateoholic.com site works ;)
But probably more familiar to most people is Amazons recommendation
system. ("Others that have liked this like this....")

----

So theres many ways of doing it, and to some extent all would be
usefull to geolocated data in some situations.
All of them, as well, have their downsides - both in terms of
bandwidth and cpu, as well as various levels of vulnerabilities to
abuse and difficulty to set up.

So I feel the POI spec shouldn't be pinned down to any specific system
for trustworthiness, but should allow the potential for any 3rd party
to start one.
I don't think this would be too hard as long as POI's can be uniquely
identifiable. Then the trustworthiness evaluation sources could either
index and create small notes next to them, or (my preference),
position those notes relative to the original POI.


-Thomas

On 15 February 2011 11:47, Jens de Smit <jens@layar.com> wrote:
> On Tue, Feb 15, 2011 at 11:19 AM, Roy Davies
> <roy.c.davies@flexstudio.co.nz> wrote:
>> Hi all,
>>
>> Interesting point.  This is something I've contemplated in my current implementation of POI, and come up with an idea built on the Maori concept of Mana.  Mana is a complex concept, but roughly means power, trust, strength, regardedness - those sorts of things, and is not something you get yourself, but is something conferred on you by your peers, family, greater family, public, etc.
>>
>> Mana can apply to people and things, and there are aspects like - an object owned by someone of high mana has in itself high mana, and your mana is the sum of the mana of you, your tribe, your family, your deeds, etc.  So a bad egg in the family can bring down everyone else's mana.
>>
>> So, what does this mean in the terms of POI - well, in my implementation, we are considering putting a level of Mana on on each point and person.  When a person of high mana creates   a point, this point has a higher mana level than a point created by someone with lower mana.  Further, points that people like, and approve of, increase in mana, and correspondingly, that person's mana increases.  Of course, the opposite is also true - if people, generally, disapprove of your POIs, then the point's mana drops, and so does yours.  A high level of mana is associated with trustworthiness in the system, and perhaps even access to certain parts of the system that others can't so easily get at.  And a low level of Mana means you might eventually get kicked out.
>>
>> Is this sensible and implementable?  Don't know yet, I'll let you when we implement it, and then how it tracks.
>
> If these relations between people and their information remain forever
> and remain being updated, I bet we could make money selling
> visualization tools. Take some psychedelic colors for the viz, get
> some people hooked on the concept so that the Mana starts flowing, and
> we have the perfect mixture between an LSD trip and Facebook.
> Guaranteed gold. Wonderful use case for video glasses too.
>
> In all seriousness, I wonder how feasible it is to keep updating
> everything. It seems to me the computational complexity explodes with
> the number of points and the number of participants. Doing on-access
> evaluation of a reasource's "Mana score" may introduce high latency,
> destroying part of the experience. Proactive updating has the risk of
> triggering a system-wide update storm. A propagation cutoff (dropping
> Mana updates once the delta gets small) seems in order to limit the
> extent of updates but that introduces the risk of introducing
> unfairness, when 10 updates of 0.1 get dropped but 1 update of 1.0
> comes through fully.
>
> Still, interesting concept.
>
> Regards,
>
> Jens
>


The information contained in this communication may be CONFIDENTIAL and is intended only for the use of the recipient(s) named above.  If you are not the intended recipient, you are hereby notified that any dissemination, distribution, or copying of this communication, or any of its contents, is strictly prohibited.  If you have received this communication in error, please notify the sender and delete/destroy the original message and any copy of it from your computer or paper files.

Received on Tuesday, 15 February 2011 14:17:22 UTC