on electronic participation, collective moderation, and voting systems Issue 2
2014-10-07
Liquid Democracy for Civic Participation – A View on LiquidFriesland
Despite the widespread dream of replacing representative democracy with some form of Liquid Democracy on a state level, LiquidFeedback has been developed for usage within political parties and not for direct use through any citizen. As inventors of the software, we have often been very reluctant when we heard about proposals to facilitate LiquidFeedback for civic participation.[Kistner2012] Nevertheless, a county in the north of Germany, the county of Friesland (Landkreis Friesland), came up with a promising plan to facilitate LiquidFeedback for a new form of civic participation. Their project carries the name “LiquidFriesland” and has been put to practice since November 2012.
There have been debates about whether LiquidFriesland shall be considered a success or not. While there was an unanimous decision in Friesland's county council to launch the system,[Kreistag2012] from the opposition it was claimed that at least 100 continuously engaging citizens should be the benchmark for measuring LiquidFriesland's success.[Drehkopf2012] But how big is the participation quota of LiquidFriesland in reality? And how big must the participation quota be in order to call a civic participation project being “successful”? Before these questions can be answered, we should take a look at the goals of LiquidFriesland.
Additional channel
LiquidFriesland was meant as an additional channel to give citizens the opportunity to be heard in the political process.[PressRelease][Leaflet] While LiquidFeedback would allow binding decisions of its participants, the results of LiquidFriesland are not meant as binding decision but as input to the county council.[Projekt] Binding decisions where all citizens are represented were neither intended nor necessary for this level of citizen participation. Anyone who is not able or not willing to engage in the online discourse may still revert to old-style signature lists as a petition to the politicans. Also here, a representative quantification is neither intended nor necessary.
Representation, quantification – it's complicated
LiquidFeedback's voting and quantification mechanisms are, in case of LiquidFriesland, only suitable to give qualitative feedback (e. g. “there have been some opposing voices”) instead of quantitative feedback (e. g. “a majority of our citizens is opposing this proposal”) in regard of the county's whole population. [Note: Of course, LiquidFeedback still produces quantitative feedback, but due to the expected selective participation these results only give a limited inference on the citizens' opinion.] The quantification process of LiquidFeedback, however, is still meaningful. LiquidFeedback's vote and support counting mechanisms and its algorithms for minority protection (Harmonic Weighting and Proportional Runoff) ensure that a group inside the system may not appear bigger by posting more often. The representation of people solely depends on whether they participate or not, and not on their degree of “noisiness”.[PLF, section 4.10] This is a major advantage when comparing LiquidFriesland to other online discussion systems.
Since it is ensured that only eligible participants may get one account (and not more than one account), the feedback for the politicians is quantitative in such way that politicians get an absolute count of verified identities of people who are in favor of a proposal and an absolute count of verified identities of people who are against a proposal. Not only a single group of citizens is heard, but also those citizens who are against a proposal or who suggest alternative proposals.
Data hunting
Even if the number of participants may not be the right benchmark for evaluating a civic participation system,[PLF, subsection 6.1.5] and even if the sole existence of such a system can change the attitude of both citizens and politicans independently of the actual usage of the system,[PLF, p.136] we want to take a closer look at the number of participants in LiquidFriesland.
Unfortunately, the access to raw data is limited due to Friesland's decision to not publish the ballots of their participants (the ballots are only visible for registered participants from Friesland). We will thus have a look at the publicly available information, which is (a) the aggregated voter counts in the system itself [LFrSystem], and (b) statistics published by the county [Evaluation].
Gathering the publicly available information at https://www.liquid-friesland.de/lf/index/index.html?tab=closed we find the following vote counts:
![Verwaltungsverfahren = administration instigated issues,
Bürgerverfahren = citizen instigated issues.
Date refers to the date when the issue was closed/finished.
For issues where at least one initiative has passed the second supporter quorum, the numbers of approvals, disapprovals, and explicit abstentions in the final voting are given for the preferred initiative.
(Note: The sum of approvals, disapprovals, and explicit abstentions is the same for all initiatives within an issue.)
For issues where no initiative has passed the first or second supporter quorum, the numbers of supporters, potential supporters, and remaining people in the reference population[PLF, section 4.9] are given for the most supported initiative.
2012-11-30 Verwaltungsverfahren #13 with 7 approvals, 0 disapprovals, 0 explicit abstentions;
2012-12-07 Verwaltungsverfahren #8 with 18 approvals, 7 disapprovals, 1 explicit abstension;
2012-12-07 Verwaltungsverfahren #18 with 11 approvals, 2 disapprovals, 1 explicit abstention;
2012-12-07 Verwaltungsverfahren #6 with 10 approvals, 5 disapprovals, 1 explicit abstention;
2012-12-08 Verwaltungsverfahren #1 with 24 approvals, 18 disapprovals, 6 explicit abstentions for the preferred initiative;
2012-12-11 Bürgerverfahren #2 with 4 approvals, 20 disapprovals, 3 explicit abstentions;
2012-12-11 Bürgerverfahren #3 with 8 approvals, 4 disapprovals, 4 explicit abstentions for the preferred initiative;
2012-12-11 Bürgerverfahren #4 with 22 approvals, 6 disapprovals, 9 explicit abstentions for the preferred initiative;
2012-12-11 Bürgerverfahren #5 with 12 approvals, 3 disapprovals, 4 explicit abstentions;
2012-12-15 Bürgerverfahren #9 with 20 approvals, 20 disapprovals, 10 explicit abstentions for the preferred initiative;
2012-12-15 Bürgerverfahren #10 with 10 approvals, 4 disapprovals, 1 explicit abstention;
2012-12-16 Bürgerverfahren #11 with 19 approvals, 4 disapprovals, 2 explicit abstentions;
2012-12-16 Bürgerverfahren #12 with 12 approvals, 2 disapprovals, 2 explicit abstentions;
2012-12-26 Bürgerverfahren #14 with 10 approvals, 3 disapprovals, 1 explicit abstention;
2012-12-27 Bürgerverfahren #15 with 18 approvals, 2 disapprovals, 2 explicit abstensions;
2012-12-29 Bürgerverfahren #16 with 14 approvals, 3 disapprovals, 3 explicit abstentions;
2013-01-02 Bürgerverfahren #17 with 14 approvals, 13 disapprovals, 4 explicit abstentions;
2013-01-15 Bürgerverfahren #19 with 2 supporters, 0 potential supporters, 39 non-supporters in reference population;
2013-02-10 Bürgerverfahren #26 with 4 supporters, 0 potential supporters, 48 non-supporters in reference population;
2013-02-12 Bürgerverfahren #27 with 3 supporters, 0 potential supporters, 50 non-supporters in reference population;
2013-02-15 Bürgerverfahren #20 with 18 approvals, 2 disapprovals, 1 explicit abstention;
2013-03-01 Verwaltungsverfahren #21 with 33 approvals, 2 disapprovals, 1 explicit abstention;
2013-03-07 Bürgerverfahren #22 with 16 approvals, 1 disapproval, 3 explicit abstensions;
2013-03-08 Bürgerverfahren #24 with 24 approvals, 0 disapprovals, 4 explicit abstentions;
2013-03-09 Bürgerverfahren #23 with 12 approvals, 3 disapprovals, 2 explicit abstentions;
2013-03-09 Bürgerverfahren #25 with 12 approvals, 1 disapproval, 3 explicit abstentions;
2013-03-13 Bürgerverfahren #28 with 3 approvals, 0 disapprovals, 2 explicit abstentions;
2013-03-16 Bürgerverfahren #30 with 12 approvals, 2 disapprovals, 1 explicit abstention;
2013-03-17 Bürgerverfahren #31 with 23 approvals, 0 disapprovals, 2 explicit abstentions;
2013-03-17 Bürgerverfahren #29 with 7 approvals, 5 disapprovals, 0 explicit abstentions;
2013-03-19 Bürgerverfahren #32 with 13 approvals, 0 disapprovals, 2 explicit abstentions;
2013-03-22 Bürgerverfahren #36 with 5 supporters, 0 potential supporters, 54 non-supporters in reference population;
2013-03-27 Bürgerverfahren #38 with 4 supporters, 0 potential supporters, 47 non-supporters in reference population;
2013-04-04 Bürgerverfahren #33 with 13 approvals, 0 disapprovals, 1 explicit abstention;
2013-04-11 Bürgerverfahren #40 revoked by initiator;
2013-04-16 Bürgerverfahren #34 with 7 approvals, 3 disapprovals, 0 explicit abstentions;
2013-04-16 Bürgerverfahren #35 with 7 approvals, 1 disapproval, 3 explicit abstentions;
2013-04-18 Bürgerverfahren #41 with 3 supporters, 0 potential supporters, 58 non-supporters in reference population;
2013-04-23 Bürgerverfahren #42 with 1 supporter, 0 potential supporters, 61 non-supporters in reference population;
2013-04-24 Bürgerverfahren #37 with 8 approvals, 1 disapproval, 5 explicit abstentions;
2013-05-14 Bürgerverfahren #39 with 13 approvals, 0 disapprovals, 3 explicit abstentions;
2013-06-09 Bürgerverfahren #43 with 30 approvals, 1 disapproval, 6 explicit abstentions;
2013-06-13 Bürgerverfahren #44 with 15 approvals, 3 diaapprovals, 1 explicit abstention;
2013-06-14 Bürgerverfahren #48 with 5 supporters, 0 potential supporters, 49 non-supporters in reference population;
2013-06-17 Verwaltungsverfahren #45 with 8 approvals, 1 disapproval, 2 explicit abstentions for the preferred initiative;
2013-06-17 Verwaltungsverfahren #46 with 33 approvals, 1 disapproval, 6 explicit abstentions;
2013-06-29 Bürgerverfahren #53 revoked by initiator;
2013-06-29 Bürgerverfahren #54 revoked by initiator;
2013-07-05 Bürgerverfahren #52 revoked by initiators;
2013-07-09 Bürgerverfahren #47 with 19 approvals, 1 disapproval, 3 explicit abstentions;
2013-07-27 Bürgerverfahren #49 with 5 approvals, 3 disapprovals, 1 explicit abstentions;
2013-07-29 Bürgerverfahren #51 with 21 approvals, 3 disapprovals, 1 explicit abstention;
2013-07-31 Verwaltungsverfahren #50 with 21 approvals, 5 disapprovals, 8 explicit abstentions;
2013-08-04 Bürgerverfahren #55 with 17 approvals, 0 disapprovals, 1 explicit abstention;
2013-08-10 Bürgerverfahren #56 canceled by administrative intervention;
2013-08-16 Bürgerverfahren #59 revoked by initiator;
2013-09-11 Bürgerverfahren #58 with 17 approvals, 0 disapprovals, 1 explicit abstention;
2013-09-18 Bürgerverfahren #57 with 17 approvals, 0 disapprovals, 1 explicit abstention for the preferred initiative;
2013-09-24 Bürgerverfahren #61 with 24 approvals, 0 disapprovals, 3 explicit abstentions;
2013-09-24 Bürgerverfahren #60 with 13 approvals, 2 disapprovals, 2 explicit abstentions;
2013-10-10 Bürgerverfahren #64 with 4 supporters, 1 potential supporter, 65 non-supporters in reference population;
2013-10-12 Bürgerverfahren #65 with 5 supporters, 0 potential supporters, 53 non-supporters in reference population for most supported initiative;
2013-10-16 Bürgerverfahren #63 with 11 approvals, 0 disapprovals, 1 explicit abstention;
2013-10-30 Verwaltungsverfahren #62 with 8 approvals, 6 disapprovals, 5 explicit abstentions;
2013-11-07 Bürgerverfahren #66 with 5 approvals, 12 disapprovals, 0 explicit abstentions;
2013-11-08 Bürgerverfahren #67 with 13 approvals, 1 disapproval, 2 explicit abstentions;
2013-12-11 Verwaltungsverfahren #69 with 68 approvals, 12 disapprovals, 24 explicit abstentions;
2013-12-26 Bürgerverfahren #68 with 18 approvals, 2 disapprovals, 1 explicit abstention;
2014-01-08 Bürgerverfahren #71 revoked by initiator;
2014-01-14 Bürgerverfahren #70 with 26 approvals, 4 disapprovals, 2 explicit abstentions;
2014-02-16 Bürgerverfahren #72 with 9 approvals, 3 disapprovals, 2 explicit abstentions;
2014-02-16 Bürgerverfahren #73 with 7 approvals, 5 disapprovals, 1 explicit abstention;
2014-02-20 Verwaltungsverfahren #74 with 9 approvals, 3 disapprovals, 2 explicit abstentions for the preferred initiative;
2014-03-08 Bürgerverfahren #75 with 12 approvals, 4 disapprovals, 2 explicit abstentions;
2014-04-22 Bürgerverfahren #76 with 4 approvals, 0 disapprovals, 0 explicit abstentions;
2014-05-30 Bürgerverfahren #77 with 3 supporters, 0 potential supporters, 66 non-supporters in reference population;
2014-06-22 Bürgerverfahren #79 with 18 approvals, 0 disapprovals, 1 explicit abstention;
2014-06-23 Bürgerverfahren #78 with 13 approvals, 1 disapprovals, 0 explicit abstentions.](figures/figure-07-01.svg)
We can consider approvals, disapprovals, and explicit abstentions as active participation in the system. In case of issues that didn't enter voting, we could still count supporters and potential supporters as active participation, but these numbers are misleading because people opposing all initiatives are missing here (the reference population also contains participants subscribed to a subject area[PLF, section 4.9] who are not actively participating). Therefore, we only consider that issues where at least one initiative passed the second supporter quorum (i. e. where at least one initiative entered the final voting procedure), and we add approvals, disapprovals, and explicit abstentions to determine the count of active participants during voting. We should note that the actual count of participants may be higher, as people could have engaged in the discussion process but refrained from voting.
Method A: average participants per issue
There are different ways to interpret participation numbers. The first idea coming to one's mind is to calculate the average number of participants per issue for a given time frame. While this method seems to be suggesting itself, it is not suitable as a general indicator of participation: additional issues (hence extra participation) could lower the average value.
Method B: maximum number of participants per issue
A more useful value might be the maximum number of participants per issue within a given time frame (see Figure 3). But this value still suffers another problem: the more people are interested in the same issue, the higher the number gets. To give an example: assuming there are 100 people participating per quarter of the year, then the maximum number of participants per issue during that quarter could be 100, but it could also be 10 if all participants split up into disjoint groups of 10 persons each, who are engaging in different issues. Thus, neither the maximum number of participants per issue in a given time frame is suitable as a general indicator for the total amount of participation.
Method C: counting every active participant
Arguably the best measurement for participation quota is the number of people showing activities within a given time frame independently of the particular issues they deal with. Unfortunately, the values in Figure 2 are not containing enough information to deduce that numbers. Just by counting the participation of each issue independently, we have no idea about the disjointness of the participants involved in each issue. The county of Friesland only published numbers aggregated over the whole run time of the system: 458 participants were active in at least one issue since startup of the system, 183 participants were active in at least two issues, and 116 participants were active in at least three issues.[Evaluation] The evaluation report where these numbers are published unfortunately doesn't say anything about the development of these numbers over time (e. g. participants during a month or during the quarter of a year).
Method D: total number of votes
Another number that may be of interest is the sum of cast votes for a given time frame (see Figure 4). Here, not only the number of participants is taken into account, but also the intensity of each person's engagement.
0.1 percent
Keeping the limitations of these numbers in mind, we may still deduce that in the 4th quarter of 2013 at least 104 participants were active in the system (see Figure 3), and in the 2nd quarter 2014 at most 37 participants were actively voting in the system (see Figure 4). Nevertheless, considering the last year, more than 100 participants used the system. Given Friesland's population of approximately 100,000 inhabitants, this is about 0.1% of the total population.
While this number looks pretty small, let's compare it with another citizen participation system in Germany on the federal level: the e-petition system of the German Federal Parliament (Deutscher Bundestag).[ePetition] The German Parliament requires a quorum of 50,000 supporter votes for a petition to be discussed in the petition committee.[ePetitionQuorum] A quorum of 50,000 is approximately 0.062% of the total population. According to the published list of petitions within the system, only 18 petitions have passed this quorum.[ePetitionList]
With this considered, the participation quota in LiquidFriesland is at least in the same order of magnitude as the e-petition system of the German Federal Parliament. Does that mean LiquidFriesland is just another electronic petition system?
LiquidFriesland vs. e-petition systems
There are a number of differences between LiquidFriesland and, for example, the e-petition system of the German Federal Parliament. First of all, LiquidFriesland allows development of multiple competing initiatives. Each participant may freely post any initiative without having it reviewed or merged by a commission, while this is the case in the e-petition system of the German Federal Parliament.
Contrary to the e-petition system, in LiquidFeedback (and thus in LiquidFriesland) it is always possible to post alternative proposals and counter arguments independently of a request commission or petition committee. Those proposals and arguments are quantified in regard of their supporters and presented in a way that minorities may present their point of view in an adequate way.[PLF, chapter 4]
Utilizing LiquidFeedback's design principles, LiquidFriesland has a lot of desirable properties that classical petition processes cannot offer. It's potential, however, is by far not used yet.
Degree of bindingness
The county council in Friesland (Kreistag) has committed itself to put every successful initiative in LiquidFriesland to their agenda. In most cases this is, by now, the maximum possible extent of bindingness achievable. For a truly citizen driven democracy, it may be also thinkable though, that the participants of a LiquidFeedback system prepare a binding referendum (which is then decided using a secret ballot).[PLF, subsection 6.3.4]
The administration
Despite the question of improving a system's bindingness, another huge factor for the success of a participation system is the involvement of the political administration. In case of LiquidFriesland, the political administration tracks initiatives after being approved by a majority of participants and thus gives participants a feedback what happened with their initiatives in the political process.[LFrTracking]
A possible improvement here would be to open a dialogue with the participants already before the final voting ends: in case of unfeasible proposals, the administration could communicate with the participants during the discussion process and could work towards feasible and realistic proposals. The involvement of the administration should be just on an informative level: due to LiquidFeedback's collective moderation system with proportional representation, individual proposals of the citizens do not need to be filtered or merged against the will of the respective initiators.[PLF, chapter 4] Notwithstanding, the political administration should engage in the debate early enough. Letting citizens vote first and telling them later (i. e. when the citizens' vote is over) that their proposal won't be put into practice can have a devasting effect on the participants' motivation. [Note: An example where Friesland's administration failed to communicate the nonfeasibility of a proposal early enough can be found at [ShredLF], [ShredKT]].
The politicians
But not just the political administration could participate more actively in the process: the same holds for politicians. If politicians entered the system, the discussion process would be enriched with a lot of political expertise. The politicians in Friesland, however (at least those that are in the county council), have deliberately decided to not participate in the system.[Projekt, page 8] They want to grant “ordinary citizens” a pole position in the discussion process. While this may appear noble at first, it also relieves politicians from facing their citizens in a fair discussion process: the citizens debate on their own, and in the end the politicians either approve it, or they bring up arguments why the citizens' proposals are unfeasible (after the discussion of the citizens has ended).
A better approach would be to carry the parliamentary processes into the participation system itself. Proposals that are debated on in the parliament (or county council) should be posted in the LiquidFeedback system in due time (e. g. at least a month before a decision has to be made) to give citizens enough time to evaluate them, optionally publish counterproposals and to discuss pro- and contra arguments by posting their own initiatives. Politicans and citizens would debate at eye level.
Conclusion
For now, these ideas are just dreams. LiquidFriesland has made the first steps and is certainly a success regarding previous experiences with citizen participation. But it is important to move forward and take further steps to empower all human beings to have a more direct way of influencing politics in a fair and transparent process.