Upload
pietro-speroni-di-fenizio
View
623
Download
1
Tags:
Embed Size (px)
DESCRIPTION
This talk was presented at the FET 11 presentation. It was not recorded there, so it has been re-recorded to present it on the internet.Mass Online deliberation, how to make it happen.Alternative Title: Procedural And Algorithmic Aspects of Mass Online DeliberationAuthors: Pietro Speroni di Fenizio, Alois, Paulin, Cyril VelikanovIf a really topical and controversial problem is proposed for public deliberation, and there are solid expectations that the results of deliberation will be at least influential if not decisive - then we can expect that really many citizens (thousands, tens of thousands...) will consider it worthwhile to join an open online deliberation over that issue, and will actively participate in it.However, such a mass online deliberation (MOD) raises several problems, related to how a very large number of users' one-to-many interactions can be coordinated and aggregated. These problems can only be solved by using a carefully designed MOD support system (MOD-SS).We analyse those problems and propose our solution to them, based on the principles of fairness and self-moderation, and using a special kind of two-parameter evaluation of participants' proposals. This makes it possible for the support system to cluster proposals according to how similar they are perceived by other participants.Our system will be implemented as a server cloud application with an open API and several alternative client applications for different use-cases.
Citation preview
PROCEDURAL & ALGORITHMIC ASPECTS
OF MASS ONLINE DELIBERATIONAn R&D project by
Pietro Speroni di Fenizio(Coimbra University, Portugal)
Alois Paulin(Faculty of IS, Novo Mesto, Slovenia)
Cyril Velikanov(Memorial Society, Moscow, Russia, &Fondation Euractiv/Politech, Brussels)
Mass online deliberation: the context
Agenda-setting
Problem or question
MASS ONLINE DELIBERATION
SYSTEM
Citizens
Govern-ment
Experts
Solution(s) or answer
(s)
Probably, also other stakeholders
Parliament Regional Council
Citizens, or other stakeholders
Deliberation system type will depend on:Number of ParticipantsNumber of ParticipantsNumber of ParticipantsNumber of Participants
<12 10-150 150-1000 >1000
What are we looking
for
One Answer(How should we do this?)
What are we looking
for
Multiple Answers
(What should we do about
this?)
MOD: Citizens’ Actions
MASS ONLINE DELIBERATION
SYSTEM
Citizens
ProposingCommenting Voting
Modifying Appraising
Manyto
Many
MOD Principles: Fairness
Every proposal should be appraised by the deliberating community according to its intrinsic value,
regardless of:- author’s personality- friends’ support- submittal time- etc
MOD Principles: Scalability
MOD Principles: Informed opinion
Uninformed opinion is of little value. Information should come from independent sources. Experts (academics?) should provide data, but not opinions
The deliberation process should be able to host thousands of contributors
MOD Principles: Economy of efforts
MOD Principles: Robustness
We value deliberative actions of well-intentioned citizens.So we must protect them from concerted disruptive actions (“mob attacks”)
Participation should be possible to everyone at leisure time – not as a full-time occupation.
MOD Principles: No External ModerationThe system should be self-moderated by participants’ own actions, with no need for external agents that can bias the result.
Moderation Unmoderated
Self- moderated
• Endless ramification of the discussions: as in an ordinary online forum
• Trolling: Open discussion on a topical issue is prone to all kinds of concerted disruptive actions (“mob attacks”)
• Claque voting: A proposal can easily win if its author is supported by his/her many “online friends”
Problems related to self-moderation
10
• Attention limit: Out of a very large “heap” of proposals or opinions, everybody will read just a few ones
• Minority voices:If the whole “heap” is presented as one ranked list, only the most supported voices will be heard
OUR SOLUTION: Random appraisal of proposals, and then clustering them according to whether they are supported mainly by the same participants
Main problem of MOD: Making sure all are heard
Our solution: random two-parameter appraisal
• Initial appraisal: Every participant’s contribution is first sent to few randomly selected participants (a “peer review”)
• Two-parameter appraisal:
quality:
“how well it is explained?”ag
reemen
t:
“do I a
gree w
ith it?
”
Quality
Rankingbetween proposals
Agreement
Clusteringbird’s-eye view:
“best” contributions for each cluster
quality
agree
ment
Aggregating Several Appraisals
Closenessbetweenproposals
Citizens&
Stakeholders
Proposals
Clusters
?ProposalsProposalsProposalsProposalsProposals
Experts
Appraising
ClustersClusters
Writing
MOD: authorship vs. confidentiality
SOLUTION: Unique registration under user-selected pseudonym with digital signature
• Confidentiality: Though, many participants would like not to disclose their real names.
• Rewarding participants: Many participants would like to be rewarded for their efforts and for their time spent in deliberation; authors would like to have proofs of their authorship.
Collaboration within online deliberation
Deliberation should comprise collaboration, otherwise it will remain fruitless.
This raises several issues:
SOLUTION: Several workgroups in parallel; every workgroup is self-governed according to one of a choice of policies
• Authorship: Who should be considered the author of a collaborative proposal?
• What incentives should be provided to convince authors to collaborate?
Collaborative Working Groups of two types:1.Clarity seeking WG:
Working Group with only people that agree with the proposals inside a cluster, with the aim to edit a single proposal that represents the whole cluster
2.Integration seeking WG: WG with people from different clusters, trying to integrate different ideas from those clusters, with the aim to elaborate a final proposal
3. Among the proposals written by each i-WG, a ranking is established using Condorcet Voting
Cluster Cluster Cluster
IntegrationWG
IntegrationWG
IntegrationWG
Clarity WG Clarity WG Clarity WG
Final Proposals
VOTING
Citizens&
Stake-holders
Proposals
Clusters
?ProposalsProposalsProposalsProposalsProposals
Experts
EvaluationsClarity WG
ClustersClarity WG
ClustersClarity WG
Integration WGIntegration WGIntegration WGFinal
ProposalsFinal
Proposals
WG participation
Votin
g
Writing
Ranked Results
1919
Realisation: MOD Cloud(engine + storage)
2020
MOD Cloud(engine + storage)
API
http://…
Realisation: Multiple Equivalent MOD Clients