Trust Metrics

From RCC 2007

In attendance and interest in the topic:

• KevinTurner - see what people are doing
• Mickki Langston - issues of trust and reputation very fundamental to community building work of OS Earth, SourceTree, etc
• RayKing - trust across wikis
• John R - how do we deal with icky stuff about trust
• Alex Malinovich - nasty issues around trust - who controlls things like identity data-bases, etc
• IsabelAnheier - new to wikis, trying to learn
• Bart Massey - interested / academic
• David Pool - news for neighbors - built on slash - with trust built in - how do you translate your expertise in one area to another community
• Keith Lofstrom - chip engineer - ICID - hardware identification; how do you use that without violating people's privacy; manage misuse
• ArthurBrock - signal to noise filtering / trust metrics
• JakeElliott - media artist, systems approach to art; social sculpture
• Ron Johnson - SysAdmin for About Us - default mode is to not trust anyone (as a sysadmin)
• Julian Blake Kongslie - (student of bart) security groups for authorization and authentication - needs to be secure method of both
• Virginia Hammon - Trust is foundation of every community; curious about application

and others who are wandering by..

http://www.jyte.com - "Just You Tell 'Em" - making claims and sharing cred

There exist a number of different algorithyms to calculate a trust metric. Examples include: Google page rank, Advogado (virgule), ad hock metric sytle with voting, etc; also PGP (Pretty Good Privacy), an email security system where people are certified, etc to send/receive email messages. (a trust community)

Question: What do we mean by trust? ("I've never seen any trust") For one, trust is an assessment. There's sincerity and competence. If what you are saying publicly matches who you are privately, that's honesty. But I wouldn't trust you to fly an airplane, because I need evidence of competence. There's an issue of sincerity.

There may be other metrics. If you have persistent identity and anyone can attach metrics to that identity (a person makes 2,000 great wiki edtis without any vandalism) so by the time you see the guy's metrics attached to his identity, you can make decisions about whether the person is trustworthy, based on the information you need and that's relevant to the relationship / interaction you have or want.

Based on this example, you may be less inclined to look over all the changes he's made because it's trustworthy - you know he's made 2,000 edits without a single vandalism. You know you can trust that person to make future edits.

The challenge is to apply the basics of trust in the digital world.

It's nice if people who aren't computer geeks can understand what's going on - and it's good to mirror what's happening in the real world to help with that.

In many situations, you have to trust the person who is trusting the next person.

Jyte is a Social experiment - with your indentifier, other people online can make claims about you, and the community can say: "yes" or "no" to validate whether the claim is true. It builds a network of what the community says is true or false.

Context of repuation - I have different repuations in various communities. How do you model that, and how do you enable that.

Brings up the issue that you'd then have multiple identities, if one repuation gets trashed.

Why shouldn't communities be able to determine what kinds of identity persistence and validation they require for their purposes? Currently, our systems do now allow that kind of flexibility.

The minute something is public on the internet, it's public forever. It's a one-way street. It's a reality of the internet today.

Move back a step - drew distinction between communities that want to have an identity anchored and those that don't. There's also a way to make it difficult to let a single person make double identities - like gold exchange accounts. It requires lots of money to open an account, but it doesn't have tons of validated personal information. How can we make it difficult to make lots of identities without it being anchored to a persistent real identity (for communities who don't want that)

Work-based identity - people work hard to create an identity, and it's important to them for it to persist.

Look at WHY you want trust so you can work backwards and create a system that gives the information you need to have that trust

About Us has a captcha mechanism that decreases over time based on legitimate use - goal is to eliminate spam.

Sharing identity across communities requires combination of two things: something like SisterCred which publishes stats about user participation in wikis and verifiable identity across sites.

Jyte has a concept based around contextual repuation kudos. Builds a network of people who are making assertions about people and the community as a whole can validate whether that is true - and other sites can use that information if they want to.

See also: MeatballWiki:WhatIsTrust