Jusletter IT

Democracy assumes equality of rights. With regard to voting, equality of rights is commonly understood and practised as the «one person — one vote» rule. In the subsequent sections we will discuss some potentially acceptable modifications to that rule; let us start however with a more general statement about democratic rights. Namely, when citizens vote for a given candidate or for a given solution, they make their choice within a restricted set of alternatives, which has been establishedwithout their participation , that is, in a basically undemocratic way.

This was not so in the ancient Athenian Republic, where every citizen had theright to propose , so that the agenda for public deliberations in the Assembly or in the courts was set by citizens («the ones whowant among those whocan ») and not by appointed magistrates. In this way, every citizen was entitled to propose e.g. a draft law. This was the most important and inalienable right of every Athenian, and now, 24 centuries later, both the citizens and the institutions in Europe and elsewhere are beginning to consider that our modern democracy needs a much larger citizens’ participation, in particular in law-making.

Participation, however, means not only to propose, but also to discuss proposals. The Athenian polity had at different times between 30 and 60 thousand citizens only, and the number of citizens eligible to deliberate on a given question in a given court was made yet several times less, by applying a system of rotation and selection by lot. How then can we, in a constituency of a few million (that of a small country or a large city) up to tens and even hundreds of millions (a big country, EC, USA,…) select a reasonable proposal, and then deliberate on it all together? Rotation or selection by lot works no more, because, unlike for the Athenians each of whom was in turn becoming «eligible» several times in his lifetime, we would wait for our turn probably a thousand years or longer.

The only way to implement an «Athenian-style democracy» in our today’s very large constituencies would be, therefore, (1) to accept any number of participants, i.e.all those «who want» ; (2) to let those participants to selectby themselves the most reasonable (or otherwise the best, or the most preferred) proposals and ideas; (3) to provide those other citizens in the constituency «who didn’t want, though could» with an assurance that in their absence the issue has been discussed and solved quite fairly, openly and «democratically», so that accepting the proposed solution would be for them reasonable enough. Indeed, to make these basic principles workable, if this is possible at all, we need some strong procedure to be applied. In the following sections we will discuss some aspects of this procedure.

In the last twenty years, overwhelming expansion of «electronic» communications via Internet has made the idea of open and direct «participative» governance both more desirable and more believable. Internet users look at such open development projects as Wikipedia, Linux and many others as paving the way to such an open participative governance. On the other side, public administrations at all levels of governance feel the need for a stronger legitimation of their actions, and more generally, of their mere existence. A concept called «e-Democracy» in Europe («Digital Democracy» in the USA) has thus come into the world, though it still remains, in our view, more a general intention than a well-defined direction.

Among the various paradigms for e-Democracy,e-Participation is the only one that really intends to serve the above-discussedright to propose , namely, the possibility for citizens of making and discussing legislative or administrative proposalsof their own («bottom-up»). Implementing appropriate e-Participation tools and their large scale deployment at various levels of governance in Europe, from municipal to the pan-European, is considered by the EU instances as one of the major priorities for the year 2020. However, the question of how it should be done is still largely debated, and various experimental realisations have never drawn more than few hundred participants into a common discussion of a given theme. For this reason, much of today’s efforts in this field are spent for such tasks as advertising and disseminating a given e-Participation tool, platform or campaign, making it more attractive, more easily mastered, and finally more populated, in order to broaden citizens’ participation in governance.

In contrast with the facts above, we are searching for solutions to a quite different question: how to achieve efficiency and effectiveness of an e-Participation project or campaign once the above-stated problem of atoo scanty participation is solved by whatever means, that is, when the number of participants grows up to a few thousand or even more? In particular, how to solve the ”crowd buzz” problem that appears in any mass deliberation, when many people start speaking without listening to each other? We are considering this problem in a specific context of massonline interaction of citizens, i.e. by using Internet. In fact, we restrict ourselves to a yet narrower context, when participants exchangethoughtful written communications, thus not orally and not by «chatting» (we will use the term «contributions », to emphasize the collaborative aspect of any effective e-Participation activity). In our view, this is the only way to escape the crowd buzz problem. We will calle-Deliberation the process of exchanging such (one-to-many, or even one-to-all) thoughtful written contributions over Internet, presumably on a specially designed Web-site or platform.

Another problem to be solved is how to maintainstability in e-Deliberation. Achieving stability means for us making it impossible or at least rather difficult for a spontaneously or intentionally organized majority of participants to impose their precipitate and thoughtless view on the whole «online assembly» («mob law» or «mob attack»), and at the same time making it equally impossible or very difficult for a small group of «influential» participants to impose their solution against largely discussed and clearly expressed majority will («oligarchic ruling»).

In this paper we propose a solution to the above-stated problems of mass e-Participation. To solve the «crowd buzz» problem, a method of mutual moderation and appraisal of participants’ contributions is introduced, which helps to promote the «best» contributions toward the top of the list, enhancing their overall visibility, and at the same time to sort contributions by grouping those of them which have been supported by generally the same group of participants. This method is only briefly described hereafter; more detailed description will appear in a separate paper.

Based on this mutual appraisal method, the second problem (that of the two-sided risk of a «mob attack» or «oligarchic ruling» to happen) can be reasonably limited, if not eliminated, by applying aweighted voting method. In brief, our solution consists in multiplying the value of everyone’s vote by that person’s «weight», so that this vote will count more (or less, depending on the weight) in calculations of totals and averages. Indeed the weights of participants are not assigned to them by any authority external to the forum, neither are they fixed forever. On the contrary, the current weight of every participant is dynamically (re)defined by the system, on the basis of various characteristics of his/her activity on the forum,as appraised by other participants in the same deliberation (and also, to some extent, in other deliberations on the same forum—this is an option to be separately studied.) Depending on the values (positive or negative) assigned to a participant’s contributions (and maybe also to his/her other actions) by other participants, his/her weight may increase or decrease. In this way, the process of assigning weights remainsopen anddynamic and, at the end, ratherdemocratic , as we will show in the following sections of the paper.

The goal to achieve is to have the weights of all participants distributed in such a way, that the total voting power of the most «heavy-weighted» or prominent participants is counterbalanced by the total voting power of a large majority of «light-weighted» participants. Note that those heavyweights are simply the most active participants who at the same time are or have been themost supported by others , while the majority of lightweights represents the least active and least prepared participants.

There is indeed a strong requirement that the distribution of weights thus achieved should remainequitable , that is, defined in accordance with appraisal actions of participants toward each other. In addition, it should beopen anddynamic , that is, giving everybody an opportunity to rise, while maintaining for them some risk to fall (in terms of their position on the forum). And finally, it should besmooth , that is, having many intermediate levels, so that a participant’s weight getting one level up or one level down does not impassion neither him/her nor others.

Summing up, our solutionshould not be less democratic than today’s democracies. Indeed we are not trying to build a «perfect democracy»; we are only trying to make a step in that direction.

Before passing on to the main theme of this paper, we will briefly describe our e-Participation model¹ . A specialised site or platform for e-Participation is provided² , on which a separate space is allocated for every new theme (problem, subject matter), so that discussions on different themes are held separately. Participants register on this forum, presumably under unique pseudonyms, and obtaindigital signature keys that are used to verify and preserve the authorship of every contribution.

Consider a situation when some topical problem has been proposed on an e-Participation forum and has attracted a large number of participants. Suppose that comprehensive expert data on that problem have been made available on the site, so that many participants have gotten acquainted with it and feel now well prepared to start discussing the problem³ .

Now every participant can send his/her own contribution, view a sorted/filtered list of contributions sent by others, view an individual contribution in a list, assign it someappraisal values , and also send a comment, which is another contribution appropriately marked and linked to the contribution being commented on. Participants can mark their contribution as aproposal (a «standalone» contribution) or as acomment (to some other contribution)⁴ .

Appraisal values may be, say, integers from -3 to +3; each numeric value has a corresponding lexical expression, say, «awful», «good», «excellent», etc. One distinctive feature of our proposed system is that a contribution is appraised according totwo distinct parameters : the level of itsquality (an intrinsic though informally defined parameter of a contribution), and the degree of the appraiser’sagreement with the contribution’s ideas. This agreement value is indeed fully subjective, as it highly depends on differences in beliefs, convictions, and other values honoured by different participants. In contrast, participants are expected to be impartial in their appraisal of others’ contributions quality; however, assessing this objectivity is indeed a very complex matter.

All comments to a contribution and all its appraisal values are made visible on the site, signed by their authors’ pseudonyms. Some system-calculatedaggregate values on a contribution may also be shown, e.g. some totals and averages of both quality values and agreement values. Also, a set ofpopularity counts can be maintained for every contribution.

A participant’sprofile is defined by some set of characteristics of his/her activity on the forum, in particular by the results of appraisal of that activity by other participants. We can of course define more than one different aggregate value for every contributing participant as we do it for individual contributions, in order to derive from these one or two «one-dimensional» characteristics (weights). The following aggregate characteristics of participants may be of particular importance:

• popularity of the participant’s contributions: how many other participants read and appraised at least one of his/her contributions;
• average quality value of the participant’s contributions, obtained by further averaging the set of average quality values of all his/her contributions, possibly weighted by each contribution’s «importance level» («proposals» being higher weighted than «comments»);
• average degree of agreement of other participants with the participant’s contributions, obtained in a similar way by further averaging the set of agreement values of all his/her contributions, possibly weighted as above.

This list can be certainly made longer. Additionally some characteristics can be defined that estimate the participant’s activity as a whole, including parameters reflecting how actively he/she reads and appraises contributions of other participants.

The above-listed (and possibly also some other) characteristics of a participant can be used by other participants for sorting contributions by their authors.

Let us try now, using the above-defined characteristics of a participant, to define theweight of his/her vote (a) when he/she appraises thequality of contributions made by others; (b) when he/she appraises the degree of his/heragreement with contributions made by others; and (c) when he/she selects (or rates) the proposals in the final list of edited proposals. In fact, it makes sense to define two different weights: the first for the case (1), and the second for the cases (2) and (3). We start with formulating a few quite simple principles:

Theweight of a participant’s vote is a positive multiplier by which the numerical value (either positive or negative) of the vote is multiplied, thus obtaining a «weighted vote» to be considered when calculating various derived characteristics of a given set of votes, e.g. the average or the total value of the set. The weight of a participant’s vote cannot be less than 1, i.e. no one can be deprived of his/her «starting» or «minimal» vote equal to 1. The weights of all participants should never exceed some predefined maximum value. This maximum weight should be considered a system parameter to be selected in large scale experiments, as it may greatly influence the system stability, its equity, and its attractiveness to participants.

The weight of a participant that is applied when he/she appraises thequality of contributions made by others (the «quality appraisal weight») should mainly depend on how thequality of his/her own contributions has been appraised by others.

In contrast, the weight of a participant that is applied when he/she appraises the degree of his/heragreement with contributions made by others (the «agreement appraisal weight») should mainly depend on the degree ofagreement with his/her own contributions, as appraised by other participants. The same weighting factor should be used when voting on a list of final edited proposals. This means that those participants who expressed a high degree of agreement with some participant’s contributions, have «strengthened» that participant’s voice, or, in other terms, have given him/her a part of their respective voting power — see more detailed discussion in the next section.

Also, any of the above weighting factors for a participant should take into account not only an average of those quality or agreement values assigned to him/her by other participants, but also some «popularity» characteristic of the participant; otherwise a participant whose contributions have been read and highly appreciated by two of his/her friends and by nobody other would obtain an unjustly high weight.

Now we can propose a method for calculating the two above-discussed weights of a participant’s vote. In both cases we start with calculating some non-normalized quantity (therating of a participant), which may grow unrestrictedly; then we derive from it a normalised weighting factor, which always remains within some predefined limits. Both weights are calculated in a similar way.

To calculate the participant’sweight in quality appraisals , we start with summarizing, for each of his/her contributions, all quality level values that have been assigned to the contribution by those participants who have already appraised it. This it the current «quality rating» of the contribution. Note that our definitions arerecursive , in a way that the values being summarized are in turn weighted by the current quality weights of the appraising participants. Then we summarize all those current ratings over all contributions of the participant, taking his/her «proposals» with a multiplying factor, say, 5 and his/her «comments» (to contributions of others) with a lower factor, say 1. This grand total is thequality rating of the participant.

Take the following example. A participant has made 2 proposals, having obtained a quality rating of 100 and 50 respectively, and also 10 comments, 9 of which obtained an average quality rating of 10 (thus making a subtotal of 90), while the 10th alone has achieved a rating of 160, because it commented on a very popular proposal and has become itself very popular. The total quality rating of the participant is thus equal to 5∙(100+50) + (90+160) = 750+250 = 1000.

The participant’s quality weight can now be defined as a function of his/her quality rating, monotone increasing from 1 to some upper limit, say 30. It may be e.g. a piecewise linear function that stays at 1 for rating values below some lower limit, then grows linearly for rating values between this lower limit and some upper limit, and finally, after reaching that upper limit of ratings, stays at its own predefined upper limit, say 30.

Our problem is however a bit more complicated, because at the beginning of a deliberation, when not many contributions have been done and appraised yet, and hence the current ratings of participants are quite low, it makes no sense to strongly discriminate their weights. On the other hand, at the end of a populous and active deliberation the ratings of many participants might have considerably grown, thus making little sense to assigning all of them a maximum weight.

Here is a simple method for solving this problem. Let the (current)volume of a deliberation be defined as the sum of the current ratings of all its participants (i.e. the grand total of all (weighted) quality values assigned to all contributions). This parameter clearly reflects the total activity of the participants, while inactive participants are not taken into account. Now let us define the (current) upper limit for participants’ weightsWmax as a function of the (current) deliberation volumeVdel . It may be another piecewise linear or other monotone increasing function giving, for exampleWmax =10 forVdel =10000, andWmax =30 forVdel =100000 or more. Specific parameters in the definition of this function may be first selected on the basis of some general considerations, and then adjusted experimentally.

Now, when the upper limitWmax of participants’ weights for a given current deliberation volumeVdel is known, we can apply e.g the following piecewise linear function to obtain the quality weights of all participants. LetRmax be the current maximum rating among all participants; then all participants rated from 0 tillRmax /10 are assigned the minimum weight 1, while the 5% of the top rated participants are assigned the maximum allowed weightWmax . All other participants are weighted according to a linear function increasing from 1 tillWmax . Indeed, all the numeric parameters here are taken more or less arbitrarily and can be changed or adjusted, as well as the nature of the functions themselves. However, we can see that the above formulated general requirements on weights can easily be satisfied.

The second weighting factor of a participant, namely theweight of agreement values assigned by him/her to contributions of others, can be defined using an algorithm similar to that proposed for calculating quality weights, by replacing everywhere quality values with agreement values. We would start with defining a second set of auxiliary parameters (Vdel ,Wmax etc) related this time to participants’ activity in appraising their degree of agreement with others’ contributions. The agreement weight of a participant would therefore depend on the degree of agreement of other participants with his/her own contributions.

Let us recall that both definitions (of the quality weight and of the agreement weight) are recursive, that is, based on the weights of all participants calculated for the set of all the previously assigned appraisal values. Note also that at the very beginning of a deliberation on a given theme, all participants have the rating 0, and their weights are all at 1. We can make however a further enhancement to our system, by taking into account the ratings of participants, which have been accumulated in the previous deliberations on various themes. A fair method for doing that would consist in considering each of those past ratings with its own weight (always less than 1), depending on thedegree of proximity of the respective past theme with the currently discussed one. How close the new theme is to each of the previously discussed ones may also be assessed by participants themselves. In this way, long-term assiduous «self-made legislators» would be appropriately rewarded.

Yet another improvement consists in assigning more value to recent contributions as compared with the older ones, by applying someaging factor , e.g. in the order of a fraction of percent per day. In this way, new participants would not remain too unfortunate, while a long inactivity period would cause a rating decrease.

A weighted voting system along the lines discussed above appears at first sight as unfair and unlikely to acquire public acceptance, which is particularly important if the decisions of an e-Participation forum are considered as more or less binding. There are however quite elementary reasons in favour of our method that are both well-founded and easy to explain. Participant X might be a recent or mostly inactive participant whose voting weight has the minimum starting value, say 1. However he might have followed the deliberations on a given theme, and have voted in support of communications posted by B and C who exposed an opinionO shared by X; also, X might not believe him/herself to be capable of expounding this opinion in such a brilliant manner as did both B and C.

When the process comes to the final voting stage, X would quite naturally vote for the solutionS elaborated in accordance with said opinionO . Thus X would add his/her modest «1» to the vote count in favour ofS , and would see B and C voting with their respective weights, say 20 and 25. Knowing that these high weights have been made up partially with his/her support, and that now these are significantly increasing the total count in favour ofS , X should not be vexed by the whole situation, which can even be understood as voting 3 times in favour ofS : one direct vote and two indirect votes through B and C who act as «representatives» of other people including X.

The whole system hence appears as a kind of «sum» of direct and representative democratic votes. We see that our procedure is qualitatively quite correct; as for its quantitative parameters, these can and should be adjusted in course of extensive live experiments, because no theoretical considerations can take into account all those yet unknown and unpractised aspects of social behaviour that may be exhibited by citizens in a quite new context of a massive e-Participation.

Imagine now that somewhere an e-Participation forum with weighted voting system aims at replacing a full-scale referendum. Should then the voting be only accessible to those citizens that took part in deliberation? Presumably not. Therefore, the forum voting should be «extended» to non-participants. But then, should we accept every «external» vote, and with the same, say, minimum weight? This would break the overall principle of counting only qualified, informed, «enlightened» opinions. The answer is torequire potential voters to be informed in order to be accepted for voting. Every citizen willing to vote on a set of proposed alternatives for a law should first read some informative materials, presumably prepared for this public stage of voting by forum participants acting as editors.

The requirement for potential voters to be informed should be endorsed by a kind of test or exam. At first this sounds strange – an exam for adult citizens willing to exercise their fundamental right of vote. Nevertheless, if we consider this requirement in the context of other cases existing today (e.g. exams for obtaining a driving license, or a hunting license, or tests of candidates on a job position in a company…), we would be much less resistant to this idea of a «voting eligibility test».

Designing and organizing such tests would not be a problem; much more difficult would be the preparation of informative materials in aconcise, comprehensive and impartial way . This is exactly the role that a deliberative law-developing forum could and should fulfil.

Hereafter we suggest another way of «incorporating» our voting model into today’s institutional context. Suppose that an open e-Participation committee on a given theme is createdad hoc top-down by the legislative body, or bottom-up trough an e-Petition. MPs may participate in the committee deliberations (over Internet) in line with other participants, having somehow more rights than «ordinary» participants. Those augmented rights may consist e.g. in some special moderator and/or editor rights, etc. However, ordinary participants havethe same right to propose an option or a solution as the MPs.

The final voting on a proposal, or for selecting one in a set of options, may be held either within the «traditional» legislative body (in such a case, virtually no legislative modification will be needed), or in an «extended legislative body» comprising both the MPs (e.g. with appropriately augmented «weight» of their votes) and the participants of said open ad hoc committee. The non-MP participants may also be assigned different voting weights depending on the level and the quality of their participation, as appraised by their peers. Establishing such an «extended legislature» indeed requires substantial changes to the current national legislations.

The concept of weighted voting as described above is not as «undemocratic» as it appears at first sight. On the contrary, it makes practically realisable the «right to propose and deliberate», which is now largely seen as one of the basic rights of every citizen. Weighted voting does it by helping to avoid the «crowd buzz», «mob attacks» and «oligarchic ruling» phenomena, which otherwise would hinder democratic and productive course of open e-Participation deliberations. It also provides for building self-regulatory forums that are both democratic, efficient and economic. Finally, it can be advocated as providing equal rights to citizens to enter a dynamic deliberation hierarchy and be promoted by their own efforts to the top level of that hierarchy.

Cyril Velikanov, MEMORIAL Society, Malyi Karetnyi per. 12, Moscow 127051 RU,and
POLITECH Institute, 67 Saint Bernard St., Brussels 1060, BE;cvelikanov@gmail.com