On 3/20/21 12:26 AM, Jason Cobb via agora-business wrote: > H. Herald, I submit the following document as a thesis in the hopes of > earning a degree. > > I note that the below document is a markdown document and references > images and text files that are not included directly its body. All such > files are included in the attached zip file. > > A rendered version of this thesis is available at [0]. > > [0]: http://randomcat.org/assessor-thesis/
Response to peer review by Aris Diff: https://github.com/randomnetcat/assessor-thesis/commit/e5c7eacacdcfa3f51d7a6f1d96101275733c2e9c # The Most Influential Agoran **Or: An excuse to do a bunch of math on all the assessment data I have** ## 0. ~~Questions I imagined being frequently asked~~ Preamble ### Where do these data come from? I've been using automation for assessments for most of the time that I've been Assessor. ### What time frame do these data cover? The entire time I've been Assessor (except for my first assessment, which I made by hand) until I was partway done with this thesis, so roughly from 2019-07-09 to 2021-03-15. This is about a year and half worth of data, which is probably enough to get a decent grasp of the stats for people who have participated continually through it. ### What are you going to do with these data? Influencers have been ruining our once-prosperous society by, uhh, influencing. I'm going to root out the evil influencers so that they can be tracked, imprisoned, and burned at the stake. I will also write this thesis about it, apparently. ### But what is influencing? For the purposes of this thesis, influencing will be defined as "exerting power to impact Agora". This can include both direct changes to the gamestate, causing officers to do work, or convincing people to take one's point of view. Because there isn't a great way to measure this directly, here I will look at several different measures in order to attempt to measure influence. ### How are the graphs generated? They're generated using Jetbrains Let's Plot. ### What about ways to influence the gamestate besides voting-related things? They exist, but they're not in the data I have. Sorry :P. ### Who did you think the biggest influencer would be before starting this? G. E is one of the players that has been around the longest, and e seems to participate significantly in most areas of the game that e can. E therefore seems likely to the biggest, most terrible influence of them all. ## 1. Proposals Proposals are necessary for nomics to function and can make significant changes on the gamestate, so clearly proposals will make the submitter more influential. ### 1.0 A Chart  ### 1.1 Submitted Proposals The simplest metric for proposals is how many were submitted by each person. Proposals that were submitted but never voted on will be ignored because they have no influence on the game, except to increase the Promotor's workload. The number of proposals that were voted on, though, increases *my* workload, which is clearly much more important. [Data in plain text](statistics/author_written.txt) | Author | Written | | :------------------------------ | -------------: | | Aris | 71 | | Jason | 65 | | G. | 44 | | Murphy | 28 | | R. Lee | 22 | | Alexis | 22 | | nix | 16 | | Falsifian | 15 | | Gaelan | 10 | | Publius Scribonius Scholasticus | 10 | | ATMunn | 7 | | Bernie | 6 | | twg | 5 | | Trigon | 4 | | Jacob Arduino | 3 | | omd | 3 | | Baron von Vanderham | 3 | | D. Margaux | 2 | | Warrigal | 2 | | Cuddle Beam | 2 | | grok | 1 | | Telnaior | 1 | It appears that Aris and Jason are in a relatively close first and second for number of proposals submitted, with G. coming in as a not-too-distant third. They are followed by most of the people who have been relatively active throughout the entire time I've been Assessor, with a few notable exceptions. First is Alexis, who managed to write 22 proposals in from 2020-01-06 to 2020-05-03, or just over 5 per month, reaching a tie for 5th place. These data also show people who submit relatively few proposals: Trigon, omd, and Cuddle Beam. Luckily for them, they are not dirty influencers by this metric and are, for now, safe. **MOST INFLUENTIAL: Aris** ### 1.2 Adopted Proposals Though all proposals have some impact on the game, for example by requiring the Promotor to distribute them, people to vote on them, and the Assessor to collect votes, proposals that are adopted have far more impact, as they actually have the chance to change the gamestate in some way, unlike all of those lazy rejected proposals. [Data in plain text](statistics/author_adopted.txt) | Author | Adopted | | :------------------------------ | -------------: | | Aris | 57 | | Jason | 56 | | G. | 38 | | Murphy | 12 | | R. Lee | 11 | | Alexis | 9 | | nix | 12 | | Falsifian | 11 | | Gaelan | 3 | | Publius Scribonius Scholasticus | 9 | | ATMunn | 4 | | Bernie | 4 | | twg | 2 | | Trigon | 1 | | Jacob Arduino | 2 | | omd | 3 | | Baron von Vanderham | 3 | | D. Margaux | 0 | | Warrigal | 2 | | Cuddle Beam | 0 | | grok | 1 | | Telnaior | 1 | [Note: this table is sorted in the same order as the previous (in order of descending proposals submitted).] These data, unfortunately, do not appear to be much more informative than the data about submitted proposals. They mostly follow the same patterns as above. This can be seen by the numbers generally decreasing down the chart (since the persons are sorted in descending order of proposals submitted). Once again, Aris and Jason are nearly tied at the top spot for proposal influence. The noticeable break in the pattern around Gaelan and PSS suggests that either Gaelan's proposals are unusually likely to be rejected or that PSS's proposals being unusually likely to be adopted. This can be further examined in the next section. **MOST INFLUENTIAL: Aris** ### 1.3 Proposal Adoption Rate Perhaps, instead of looking at absolute numbers for adoption, it would be better to look at the fraction of proposals authored that they are able to get adopted. A very high proposal adoption rate would imply that when a person endorses an idea enough to propose, people tend to agree with them, meaning they could have influenced others' opinions, thereby becoming an influencer. [Data in plain text](statistics/author_adopted_rate.txt) | Author | Adoption rate | | :------------------------------ | ------------------: | | Aris | 0.80 | | Jason | 0.86 | | G. | 0.86 | | Murphy | 0.43 | | R. Lee | 0.50 | | Alexis | 0.41 | | nix | 0.75 | | Falsifian | 0.73 | | Gaelan | 0.30 | | Publius Scribonius Scholasticus | 0.90 | | ATMunn | 0.57 | | Bernie | 0.67 | | twg | 0.40 | | Trigon | 0.25 | | Jacob Arduino | 0.67 | | omd | 1.00 | | Baron von Vanderham | 1.00 | | D. Margaux | 0.00 | | Warrigal | 1.00 | | Cuddle Beam | 0.00 | | grok | 1.00 | | Telnaior | 1.00 | These data show that the people who propose the very most (Aris, Jason, and G.) all have approximately the same proposal adoption rate (in the ballpark of 85%), with Aris lagging behind slightly. Below those three, adoption rates vary significantly. Noticeably, PSS has a higher adoption rate (90%) than the top three (and much higher than the people who have written similar numbers of proposals to em), though e has only written 10 proposals, so this doesn't indicate a real difference in influence. Below that point, the total proposal numbers again become too small to really draw conclusions, and they give very extreme rates. **MOST INFLUENTIAL: PSS** ### 1.4 Proposal Adopted Length  Clearly, not all proposals are created equal. A minor bug fix will be less influential on the gamestate than a proposal that rewrites several core rules. One metric for this is the total adopted word count of proposals by author; this is not necessarily a *good* metric, but it is nevertheless *a* metric. The proposal length likely also roughly correlates with how much work people have to do writing and reviewing it, meaning that people who have authored more words have had more influence on the other players. [Data in plain text](statistics/author_adopted_words.txt) | Author | Adopted Words | | :------------------------------ | -------------------: | | Aris | 9706 | | Jason | 6268 | | G. | 4577 | | Murphy | 1175 | | R. Lee | 612 | | Alexis | 3200 | | nix | 3383 | | Falsifian | 662 | | Gaelan | 274 | | Publius Scribonius Scholasticus | 808 | | ATMunn | 645 | | Bernie | 1370 | | twg | 219 | | Trigon | 184 | | Jacob Arduino | 77 | | omd | 617 | | Baron von Vanderham | 221 | | D. Margaux | 0 | | Warrigal | 76 | | Cuddle Beam | 0 | | grok | 85 | | Telnaior | 31 | Unlike the previous metrics, this metric has a clear leader - Aris, who has written more than 3000 more adopted words than the next closest competitor. As is probably to be expected, the amount of words adopted roughly decreases with the number of proposals submitted, though there are a few outliers. Alexis and nix have written thousands more adopted words than the people who have written similar numbers of proposals. Looking more into the specific proposals of each, the vast majority of Alexis's adopted words (2447 of them) come from a single proposal - P8354 "Statutory Instrumentation". This is a proposal that rewrites and adds many core rules, so it makes sense that it is relatively long. This proposal is also known for creating a fiendishly complicated system that was meant to be extensible but was never actually extended. The majority of nix's adopted words (1190 and 781, totaling 1971) come from two proposals - P8408 "Sets v1.4" and P8515 "adMinistration v1.1". "Sets" was an entirely new economy, and "adMinistration" changed a core part of that economy, so it makes sense for both of them to be relatively long. **MOST INFLUENTIAL: Aris** ### 1.5 Proposal Strength Margins Let the margin of a resolution to be F - AI*A, where F is the strength FOR, A is the strength AGAINST, and AI is the adoption index of the decision. This means that a non-negative margin results in adoption (except for 0 margin on an AI-1 proposal), while a negative margin results in rejection. Thus, a person whose proposals achieve higher voting strength margins will find it easier to pass proposals, making them potentially more influential in the future, and indicating that they have been more influential in the past by mustering agreement for their proposals. #### 1.5.1 Total Proposal Strength Margins  [Data in plain text](statistics/author_avg_all_strength_margin.txt) | Author | Author avg strength margin (all proposals) | | :------------------------------ | -----------------------------: | | Aris | 14.27 | | Jason | 14.46 | | G. | 20.63 | | Murphy | -9.44 | | R. Lee | -8.87 | | Alexis | -13.86 | | nix | 15.81 | | Falsifian | 7.64 | | Gaelan | -21.56 | | Publius Scribonius Scholasticus | 24.18 | | ATMunn | 7.86 | | Bernie | -1.50 | | twg | -4.06 | | Trigon | -14.00 | | Jacob Arduino | 4.25 | | omd | 24.00 | | Baron von Vanderham | 39.67 | | D. Margaux | -66.00 | | Warrigal | 16.50 | | Cuddle Beam | -34.00 | | grok | 41.00 | | Telnaior | 21.00 | These data (primarily the boxplot) show that the three people who have written the most proposals have relatively similar margins in the middle 50% of their proposals, with Aris having a somewhat wider range, and G. having a somewhat higher-margin middle 50%, suggesting eir proposals receive slightly more consensus than the other two's, although all three manage to have a positive first quartile. This suggests that the people who write the most proposals tend to be fairly good at it, usually getting their proposals to pass with a decent margin. Beyond those three, median and Q1/3 margins become significantly less positive, except for nix, who has achieved a non-negative Q1 margin, and PSS, who manages to have a higher median, mean, Q1, and Q3 than G., albeit while writing significantly fewer proposals, making em the most influential. **MOST INFLUENTIAL: PSS** #### 1.5.2 Adopted Proposal Strength Margin In addition to looking at voting strength margins over all proposals, we can also look at margins over only adopted proposals. This provides a sense of how much agreement an author is able to muster on proposals, i.e. whether they pass convincingly or squeak by. This is a more refined measure than looking at margins over all proposals, because it doesn't provide a sense of how often proposals pass, just how strongly they pass when they do. A higher value for this metric indicates social influence in building consensus for proposals.  [Data in plain text](statistics/author_avg_adopted_strength_margin.txt) | Author | Avg strength margin (adopted proposals) | | :------------------------------ | ---------------------------------: | | Aris | 24.30 | | Jason | 25.18 | | G. | 27.94 | | Murphy | 19.17 | | R. Lee | 23.20 | | Alexis | 20.22 | | nix | 28.17 | | Falsifian | 23.91 | | Gaelan | 13.67 | | Publius Scribonius Scholasticus | 30.87 | | ATMunn | 29.75 | | Bernie | 19.00 | | twg | 29.50 | | Trigon | 37.00 | | Jacob Arduino | 24.00 | | omd | 24.00 | | Baron von Vanderham | 39.67 | | Warrigal | 16.50 | | grok | 41.00 | | Telnaior | 21.00 | Comparing this to the previous section shows that it differs far more for authors with lower proposal adoption rates, which makes sense -- they have fewer data points that will be removed. For instance, with the three most common authors (who all have relatively high adoption rates), the relative values are the same, but the plots are just shifted upwards. The next three most common authors are much closer to the previous three than they were over all proposals. PSS is once again the most influential, achieving the highest mean margin (of those that have written a non-trivial number of proposals). The proposal with the highest strength margin is P8507 by ATMunn, with a margin of 58: ``` PROPOSAL 8507 (Happy Belated Birthday v3) AUTHOR: ATMunn CLASS: ORDINARY CHAMBER: PARTICIPATION SPONSORED: YES FOR (14): ATMunn&, Aris, Baron von Vanderham, D. Margaux, Falsifian, G.*, Gaelan@, Jason%, Publius Scribonius Scholasticus&, Telnaior, Trigon, nix%, sukil, twg AGAINST (0): PRESENT (0): BALLOTS: 14 AI (F/A): 58/0 (AI=1.0) POPULARITY: 1.000 OUTCOME: ADOPTED ``` ``` ID: 8507 Title: Happy Belated Birthday v3 Adoption index: 1.0 Author: ATMunn Co-authors: G., Falsifian Amend Rule 2585, "Birthday Gifts", by deleting the sentence "Every time it is a player's Agoran Birthday, each of the other players CAN once grant em 3 coins by announcement.", and inserting the following paragraph in its place: During a player's Agoran Birthday and the 7 days following, each other player CAN, once, grant em X coins, where X is 3 if it is actually the day of the player's birthday, and 2 otherwise. [This simplifies the rule change and fixes the bugs mentioned by Falsifian. There's no real reason to prevent players from granting the belated birthday gift just because people already granted the birthday gift.] ``` It makes sense that something like this has the highest margin. It's a simple proposal with a reasonable change to a low power rule; what's not to like? **MOST INFLUENTIAL: PSS** #### 1.5.3 Rejected Proposal Strength Margin The strength margins for rejected proposals by each author show how much an author's proposals fail by when they do fail. A person who has margins closer to zero (indicating the proposals were closer to passing) will be deemed to be more influential, as it means they were generally closer to getting their failed proposals to pass, once again indicating (slightly) more ability to build social consensus than those with more negative margins.  [Data in plain text](statistics/author_avg_rejected_strength_margin.txt) | Author | Avg strength margin (rejected proposals) | | :------------------------------ | ----------------------------------: | | Aris | -26.50 | | Jason | -49.07 | | G. | -15.00 | | Murphy | -32.19 | | R. Lee | -34.29 | | Alexis | -37.46 | | nix | -21.25 | | Falsifian | -26.48 | | Gaelan | -53.65 | | Publius Scribonius Scholasticus | -36.00 | | ATMunn | -21.33 | | Bernie | -42.50 | | twg | -26.43 | | Trigon | -31.00 | | Jacob Arduino | -48.00 | | D. Margaux | -66.00 | | Cuddle Beam | -34.00 | Once again, the changes seem to depend on what fraction of a person's proposals were adopted, but the graph maintains the same basic patterns as the graph over all proposals (except for the top three, where Jason's bar has shifted far towards negative infinity, and G.'s has shifted towards zero, both relative to the other bars around them). Other than that, some bars (notably Jason and Gaelan) have shifted down as low data points that were outliers when part of the entire set of resolutions ceased to be outliers, elongating the box and whiskers. Looking at the boxplot and at the means, there is no question. G. is the author that is most able to get eir failed proposals to be closer to passing. The proposal with the lowest strength margin was P8237 by Jason, with a margin of -120: ``` PROPOSAL 8237 (Repairing Defeated Spaceships v3) AUTHOR: Jason FOR (0): AGAINST (13): ATMunn, Aris, Bernie, Falsifian, G.$, Gaelan, Jacob Arduino, Jason, Murphy, Rance, Trigon, o, twg PRESENT (0): BALLOTS: 13 AI (F/A): 0/40 (AI=3.0) POPULARITY: -1.000 OUTCOME: REJECTED ``` This proposal made a change to a subgame that was previously repealed (but the proposal still went up for a vote after that because it failed quorum before), so it was unanimously voted against. However, this proposal should really only have received a margin of -40 because its adoption index was 1.0 at submission, but was ratified at 3.0. The proposal with the next lowest margin was P8512 by Gaelan, with a margin of -99: ``` PROPOSAL 8512 (Lime Debait) AUTHOR: Gaelan CLASS: DEMOCRATIC SPONSORED: YES FOR (0): AGAINST (11): Aris, Baron von Vanderham, D. Margaux, Falsifian, G.$, Gaelan@, Jason, Publius Scribonius Scholasticus, Telnaior, nix, twg PRESENT (3): ATMunn, Trigon, sukil BALLOTS: 14 AI (F/A): 0/33 (AI=3.0) POPULARITY: -0.786 OUTCOME: REJECTED ``` The proposal itself: ``` ID: 8512 Title: Lime Debait Adoption index: 3.0 Author: Gaelan Co-authors: Jason In Rule 2438 "Ribbons", replace the paragraph beginning "Lime," replace "three or more proposals" with "three or more proposals with different authors". ``` At first glance, this looks like a fine proposal, and it would be strange for it to receive such a consensus against. However, during the voting period, PSS cast the following vote: ``` AGAINST: This proposal is malformed and unclear on what is being replaced. ``` This resulted in several people changing their votes to AGAINST, and others voting AGAINST or endorsing Gaelan as the author, who voted AGAINST. **MOST INFLUENTIAL: G.** ### 1.6 Proposals Co-authored Directly writing and submitting proposals is not the only way to contribute to the proposal process. Many players read and provide feedback on proposal drafts, or even drafting wording that the author ends up including in the final proposal. This means that co-authoring proposals shows that a person contributed to the authoring of the proposal, thus influencing both the author and, potentially, the gamestate (if and when the proposal is adopted). [Data in plain text](statistics/coauthor_written.txt) | Person | Proposals Co-Authored | | :------------------------------ | ---------------: | | G. | 41 | | Jason | 40 | | Aris | 26 | | Falsifian | 17 | | nix | 17 | | Trigon | 16 | | twg | 15 | | Alexis | 15 | | Publius Scribonius Scholasticus | 13 | | Gaelan | 10 | | omd | 5 | | Murphy | 5 | | ais523 | 4 | | ATMunn | 3 | | R. Lee | 2 | | D. Margaux | 1 | | Oerjan | 1 | | Cuddle Beam | 1 | G. and Jason are in a very close first and second place, with G. having co-authored just one more proposal. After those two, Aris has co-authored the third most proposals, and e is followed by a batch of people who have co-authored ~15 proposals each. As e did for authoring proposals, Alexis has co-authored a surprisingly large number of proposals for the amount of time that e was a player (during the period these data cover). Another oddity is that ais523, who was never a player during the time that these data cover, has been listed as a co-author on 4 proposals (P8346, P8408, P8431, and P8460). This makes sense, given that ais523 is a watcher and has suggested proposals or contributed to their drafting, without being able to submit proposals emself. This is similarly the case for Oerjan, although e has only co-authored one proposal (P8247). A few people have co-authored more proposals than they have submitted: Falsifian, Trigon, PSS, omd, ais523, and Oerjan. This makes sense as a possibility, because being listed as a co-author generally requires much less effort than submitting an equivalent proposal (contributing to the discussion vs going through all the details by oneself). **MOST INFLUENTIAL: G.** ## 2. Voting ### Or: The other half of the data I have. ### 2.1 Number of Votes Proposals are how the players influence the game, but voting is how the players influence the outcome of proposals, meaning that a person who has cast more votes has had more influence on the gamestate than a person who has cast fewer votes. [Data in plain text](statistics/voter_votes.txt) | Voter | Vote count | | :------------------------------ | ----------: | | Jason | 371 | | Falsifian | 363 | | G. | 302 | | Aris | 283 | | twg | 232 | | Murphy | 221 | | Trigon | 209 | | ATMunn | 179 | | nix | 174 | | Publius Scribonius Scholasticus | 155 | | R. Lee | 151 | | Gaelan | 120 | | Rance | 118 | | Bernie | 105 | | Alexis | 97 | | Telnaior | 75 | | Cuddle Beam | 74 | | omd | 73 | | sukil | 63 | | Jacob Arduino | 61 | | o | 61 | | Tcbapo | 56 | | D. Margaux | 52 | | Warrigal | 42 | | Baron von Vanderham | 29 | | pikhq | 20 | | Halian | 19 | | L | 19 | | Walker | 5 | | Tarhulindur | 5 | | lucidiot | 5 | | JTAC | 4 | | Noah | 2 | | Shy Owl | 1 | The total number of decisions that could be voted on was 374. There are two voters who have voted on nearly every possible decision - Jason and Falsifian. Beneath them, the numbers drop relatively quickly. Some of this is due to not voting on every decision possible, while some of this is due to the voters not being players for the entire period that these data cover (or being zombies for part of it). **MOST INFLUENTIAL: Jason** ### 2.2 Voting Strength Voting strength determines how much a person's vote counts when evaluating outcomes. There have been multiple ways of determining strength during my time as Assessor. It was originally just a penalty for holding blots, plus a single strength point bonus for the Speaker. Then, the Ministries system was enacted, which granted voting strength bonuses to most offices for some proposals, based on the category of the proposal and of the office. After that, Extra Votes were enacted in the Sets economy. EVs provided a single strength point bonus on a single proposal. Just before writing this thesis, ministries were repealed and EVs were changed to affect strength on all proposals in their voting period, not just one, and the Speaker bonus was repealed. The only constants have been the default of three and the one point penalty for every three blots. #### 2.2.1 Maximum Voting Strength [Data in plain text](statistics/voter_strength_max.txt) | Voter | Maximum voting strength achieved | | :------------------------------ | -----------------: | | Jason | 10 | | Falsifian | 6 | | G. | 8 | | Aris | 7 | | twg | 9 | | Murphy | 5 | | Trigon | 10 | | ATMunn | 8 | | nix | 8 | | Publius Scribonius Scholasticus | 7 | | R. Lee | 6 | | Gaelan | 3 | | Rance | 3 | | Bernie | 3 | | Alexis | 5 | | Telnaior | 3 | | Cuddle Beam | 6 | | omd | 5 | | sukil | 3 | | Jacob Arduino | 3 | | o | 3 | | Tcbapo | 3 | | D. Margaux | 3 | | Warrigal | 3 | | Baron von Vanderham | 3 | | pikhq | 3 | | Halian | 3 | | L | 3 | | Walker | 3 | | Tarhulindur | 3 | | lucidiot | 3 | | JTAC | 5 | | Noah | 3 | | Shy Owl | 3 | [Note: the table is sorted so that the people with the most votes appear first.] Looking at the table, it appears that the highest voting strengths tend to come from the people who vote relatively frequently. This makes sense -- they are participating in the game more, so they are more likely to do things that increase their voting strength. This was a significant effect with ministries, since officers tend (ideally) to be fairly active, and they received voting strength bonuses for their work. Notably, two people have achieved a voting strength of 10, over three times the default strength, and the maximum that has been recorded over this time period. Trigon has achieved this 5 times, each time as follows: ``` 3 | Initial 4 | Bonus of 1 for holding Speaker 6 | Bonus of 2 for Speaker's interest in Economy 8 | Bonus of 2 for Treasuror's interest in Economy 10 | Bonus of 2 for Treasuror's interest in Economy ``` Yes, Treasuror receiving two interest bonuses is intentional. This was done in order to compensate the Treasuror for eir hard work, and because no other interest seemed to fit eir role. Speaker received an interest bonus on ordinary proposals (those which ministry bonuses affect) in addition to eir global one-point bonus. Jason reached a voting strength of 10 only once: ``` 3 | Initial 4 | Bonus of 1 for holding Speaker 6 | Bonus of 2 for Assessor's interest in Legislation 8 | Bonus of 2 for Rulekeepor's interest in Legislation 10 | Bonus of 2 for Speaker's interest in Legislation ``` This is similar to Trigon's, but using two different office interests instead of a single double office interest. **MOST INFLUENTIAL: Trigon** #### 2.2.2 Minimum Voting Strength [Data in plain text](statistics/voter_strength_min.txt) | Voter | Minimum voting strength achieved | | :------------------------------ | -----------------: | | Jason | 3 | | Falsifian | 3 | | G. | 3 | | Aris | 3 | | twg | 1 | | Murphy | 2 | | Trigon | 3 | | ATMunn | 3 | | nix | 3 | | Publius Scribonius Scholasticus | 3 | | R. Lee | 0 | | Rance | 3 | | Gaelan | 2 | | Bernie | 3 | | Alexis | 3 | | Telnaior | 3 | | omd | 3 | | Cuddle Beam | 3 | | sukil | 3 | | Jacob Arduino | 3 | | o | 3 | | Tcbapo | 3 | | D. Margaux | 3 | | Warrigal | 3 | | Baron von Vanderham | 3 | | pikhq | 3 | | Halian | 3 | | L | 3 | | Walker | 3 | | Tarhulindur | 3 | | lucidiot | 3 | | JTAC | 3 | | Shy Owl | 3 | Unfortunately, this is not particularly illuminating. This is basically a table of who has ever had enough blots to receive a penalty (without any bonuses that would counter that effect). **MOST INFLUENTIAL: nobody, really** #### 2.2.3 Average Voting Strength [Data in plain text](statistics/voter_strength_avg.txt) | Voter | Average voting strength | | :------------------------------ | -----------------: | | Jason | 3.67 | | Falsifian | 3.18 | | G. | 4.11 | | Aris | 3.58 | | twg | 3.14 | | Murphy | 3.09 | | Trigon | 3.95 | | ATMunn | 3.49 | | nix | 3.59 | | Publius Scribonius Scholasticus | 3.79 | | R. Lee | 2.59 | | Gaelan | 2.80 | | Rance | 3.00 | | Bernie | 3.00 | | Alexis | 3.29 | | Telnaior | 3.00 | | Cuddle Beam | 3.12 | | omd | 3.11 | | sukil | 3.00 | | Jacob Arduino | 3.00 | | o | 3.00 | | Tcbapo | 3.00 | | D. Margaux | 3.00 | | Warrigal | 3.00 | | Baron von Vanderham | 3.00 | | pikhq | 3.00 | | Halian | 3.00 | | L | 3.00 | | Walker | 3.00 | | Tarhulindur | 3.00 | | lucidiot | 3.00 | | JTAC | 3.50 | | Noah | 3.00 | | Shy Owl | 3.00 | Here, it is again clear that people who have voted more tend to have higher voting strengths. The people with the highest average voting strengths are G., Trigon. G. and Trigon have both been officers for a long time, so their voting strengths make sense based on Ministries. G. is the only person who has managed to have an average voting strength of above 4, an entire point above the default. **MOST INFLUENTIAL: G.** ### 2.3 Endorsements Endorsements are votes that evaluate to the vote of someone else (here, the endorsee). This gives the endorsee more influence over the outcome than they otherwise would have.  [Data in plain text](statistics/voter_endorsee_all_times.txt) | Voter | Times endorsee | | :------------------------------ | -----------------------: | | Jason | 239 | | Falsifian | 74 | | G. | 159 | | Aris | 82 | | twg | 134 | | Murphy | 31 | | Trigon | 27 | | ATMunn | 3 | | nix | 11 | | Publius Scribonius Scholasticus | 16 | | R. Lee | 73 | | Gaelan | 2 | | Rance | 0 | | Bernie | 0 | | Alexis | 13 | | Telnaior | 1 | | Cuddle Beam | 0 | | omd | 3 | | sukil | 0 | | Jacob Arduino | 0 | | o | 0 | | Tcbapo | 0 | | D. Margaux | 22 | | Warrigal | 0 | | Baron von Vanderham | 3 | | pikhq | 0 | | Halian | 0 | | L | 0 | | Walker | 0 | | Tarhulindur | 0 | | lucidiot | 0 | | JTAC | 0 | | Noah | 0 | | Shy Owl | 0 | As is usual, the numbers decrease down the list, indicating that more active people are endorsed more. Once again, this makes sense, as people who vote more are more active, and thus are more likely to be trusted, more likely to be author proposals (and endorsing the author is relatively common), and, most importantly, more likely to have controlled a zombie at some point. "Endorse <master> on all proposals" is an extremely common vote for zombies, which means that the endorsement counts are artificially inflated for people who have ever controlled a zombie. This is the case for at least Jason, Falsifian, G., Aris, twg, R. Lee, and D. Margaux. Overall, Jason, G., and twg have the most endorsements. **MOST INFLUENTIAL: Jason** ### 2.4 Non-total Endorsements As a bureaucratic efficiency, I have a way to automatically cast the same vote on all proposals within an assessment. The most common use of this has been zombies endorsing masters, but there are occasionally things like "FOR on all" or "endorse the author on all". Since the latter is extremely rare (happening only twice in my data), it can be safely ignored for this analysis. If I recall correctly, there were a few other instances of non-zombie endorsements on all decisions in an assessment, but they are infrequent enough that this analysis is not invalidated.  [Data in plain text](statistics/voter_endorsee_non_total_times.txt) | Voter | Times endorsee (non-total) | | :------------------------------ | -----------------------------: | | Jason | 75 | | Falsifian | 17 | | G. | 75 | | Aris | 58 | | twg | 11 | | Murphy | 11 | | Trigon | 2 | | ATMunn | 3 | | nix | 11 | | Publius Scribonius Scholasticus | 16 | | R. Lee | 5 | | Gaelan | 2 | | Rance | 0 | | Bernie | 0 | | Alexis | 13 | | Telnaior | 1 | | Cuddle Beam | 0 | | omd | 3 | | sukil | 0 | | Jacob Arduino | 0 | | o | 0 | | Tcbapo | 0 | | D. Margaux | 3 | | Warrigal | 0 | | Baron von Vanderham | 3 | | pikhq | 0 | | Halian | 0 | | L | 0 | | Walker | 0 | | Tarhulindur | 0 | | lucidiot | 0 | | JTAC | 0 | | Noah | 0 | | Shy Owl | 0 | Here, we see that several people with many endorsements have a much smaller number of non-total endorsements: Jason, Falsifian, G., Aris, twg, Murphy, Trigon, and R. Lee. This shows that the majority of their endorsements were not organic, which means that, though it increases the endorsee's influence over the gamestate, they do not indicate influence over fellow (active) Agorans. Under this count, Jason, G., and Aris have the most endorsements, with twg, the former third place, dropping to 7th place. **MOST INFLUENTIAL (tied): G., Jason** ### 2.5 Determinative Votes Let a voter be determinative on a decision if e voted FOR or AGAINST on that decision and if flipping only eir vote ( while still updating endorsements and other conditional votes (at least those that I have implemented to be automatically evaluated)) to AGAINST or FOR respectively would cause the outcome of that decision to change. Note that this definition permits multiple voters to be determinative on a single decision. There a few ways a voter could be determinative (all requiring em to vote in the opposite way than the eventual result): the decision could be very close (which would likely mean many voters are determinative), e could have sufficient endorsements, or e could have sufficient voting strength.  [Data in plain text](statistics/voter_determination_times.txt) | Voter | Decisions determinative | | :------------------------------ | ------------------------: | | Jason | 84 | | Falsifian | 26 | | G. | 63 | | Aris | 39 | | twg | 24 | | Murphy | 22 | | Trigon | 27 | | ATMunn | 9 | | nix | 12 | | Publius Scribonius Scholasticus | 17 | | R. Lee | 18 | | Gaelan | 5 | | Rance | 2 | | Bernie | 3 | | Alexis | 11 | | Telnaior | 2 | | Cuddle Beam | 4 | | omd | 4 | | sukil | 1 | | Jacob Arduino | 3 | | o | 5 | | Tcbapo | 5 | | D. Margaux | 1 | | Warrigal | 0 | | Baron von Vanderham | 0 | | pikhq | 1 | | Halian | 1 | | L | 0 | | Walker | 3 | | Tarhulindur | 0 | | lucidiot | 0 | | JTAC | 0 | | Noah | 0 | | Shy Owl | 0 | These data reveal that voters are generally determinative on very few decisions. This suggests that voters largely tend to agree, which would result in a single defector not actually changing the outcome of the decision. Other than that, these data, as always, show that people who vote more tend to be more influential. G. and Jason are determinative on the most decisions, which is consistent with their high number of endorsements, and Aris was determinative on fewer decisions then them while also having fewer endorsements; this helps confirm the hypothesis that endorsements contribute to voters being determinative. **MOST INFLUENTIAL: Jason** To confirm the hypothesis about why voters are influential on decisions, we can look at a few examples of proposals with a very large or very small number of decisive voters (full data [here](statistics/proposal_determinative_voter_count.txt), note that it excludes proposals with no determinative voters). One example with 9 determinative voters is P8301: ``` PROPOSAL 8301 (Consolidated Regulatory Recordkeeping v2) AUTHOR: Aris FOR (9): Alexis, Aris, Bernie, Falsifian, Gaelan, Jason, Rance, omd, twg AGAINST (2): G.$, o PRESENT (0): BALLOTS: 11 AI (F/A): 27/7 (AI=3.0) POPULARITY: 0.636 OUTCOME: ADOPTED ``` This is just a close decision, which makes every FOR voter determinative. There are many examples of proposals like this, mainly with high AIs. The AI-1 decision with the highest number of determinative voters (6) is Proposal 8198: ``` PROPOSAL 8198 (Be gone, foul demon!) AUTHOR: Jason FOR (7): Aris, Falsifian, Jason, Murphy, Trigon, Walker, twg! AGAINST (4): G.$, R. Lee, Tarhulindur, Telnaior PRESENT (0): BALLOTS: 11 AI (F/A): 19/13 (AI=1) POPULARITY: 0.273 OUTCOME: ADOPTED ``` This helps to confirm that most decisions with many determinative voters are decisions that are closely decided. Looking at some proposals with low numbers of determinative voters should help confirm that voters can be determinative due to unusually high voting strengths or numbers of endorsements. One example with a single determinative voter (PSS) is Proposal 8471: ``` PROPOSAL 8471 (Only the PM can be arbitrary!) AUTHOR: R. Lee CLASS: ORDINARY CHAMBER: JUSTICE FOR (6): ATMunn, Cuddle Beam, G.*, R. Lee~, Tcbapo, twg AGAINST (5): Falsifian, Jason, Murphy, Publius Scribonius Scholasticus%, omd PRESENT (2): Trigon, nix% BALLOTS: 13 AI (F/A): 20/17 (AI=1.7) POPULARITY: 0.077 OUTCOME: REJECTED ``` This confirms that higher-than-normal voting strength can make a voter determinative. Another example with a single determinative voter is Proposal 8487, which had omd as its sole determinative voter: ``` PROPOSAL 8288 (Glitteral) AUTHOR: omd FOR (11): Alexis$, Aris, Bernie, Falsifian, G., Gaelan, Jason, Rance, o, omd, twg AGAINST (0): PRESENT (0): BALLOTS: 11 AI (F/A): 34/0 (AI=1.0) POPULARITY: 1.000 OUTCOME: ADOPTED [ Bernie: Endorsement of twg Falsifian: Conditional resolved: no Notice of Veto was published: Endorsement of omd G.: twg is Treasuror: Endorsement of twg Gaelan: twg is the Treasuror: Endorsement of twg Rance: Endorsement of Jason o: twg is Treasuror: Endorsement of twg twg: Endorsement of omd ] ``` This is an unusual decision because a majority of its voters ultimately endorse a single person (though most of those endorsers actually endorsed twg, who then endorsed omd). This means that omd alone ultimately controlled the outcome of the decision. This is a (rather extreme) example of how endorsements can result in a voter being determinative. ### 2.6 Result Agreement by Voter A voter "agrees" with the result of a decision if e voted FOR, and the resolution was ADOPTED; or if e voted AGAINST, and the resolution was REJECTED. Alternatively, a voter "disagrees" with the result of a decision if e voted AGAINST, and the resolution was ADOPTED; or if e voted FOR, and the resolution was REJECTED. Note that this definition does not imply a voter agrees or disagrees with the result of every decision -- if e voted PRESENT or if the decision was resolved FAILED QUORUM, e neither agrees nor disagrees with the result. Correlation is not causation, but having an abnormally high result agreement rate could indicate that a person has social ways of influencing other people to vote in agreement with em, which would make em an influencer.  [Raw agreement rate data](statistics/voter_result_agreement_rate.txt) \ [Raw disagreement rate data](statistics/voter_result_disagreement_rate.txt) | Voter | Result agreement rate | | :------------------------------ | --------------------------: | | Jason | 0.75 | | Falsifian | 0.67 | | G. | 0.67 | | Aris | 0.79 | | twg | 0.73 | | Murphy | 0.64 | | Trigon | 0.72 | | ATMunn | 0.72 | | nix | 0.69 | | Publius Scribonius Scholasticus | 0.81 | | R. Lee | 0.65 | | Gaelan | 0.73 | | Rance | 0.80 | | Bernie | 0.75 | | Alexis | 0.71 | | Telnaior | 0.49 | | Cuddle Beam | 0.78 | | omd | 0.49 | | sukil | 0.63 | | Jacob Arduino | 0.66 | | o | 0.85 | | Tcbapo | 0.66 | | D. Margaux | 0.31 | | Warrigal | 0.50 | | Baron von Vanderham | 0.90 | | pikhq | 0.80 | | Halian | 0.74 | | L | 0.58 | | Walker | 1.00 | | Tarhulindur | 0.40 | | lucidiot | 1.00 | | JTAC | 1.00 | | Noah | 0.50 | | Shy Owl | 1.00 |  The above graph shows that Agorans tend to agree a lot, with most people agreeing with proposal outcomes upwards of two- thirds of the time. The results most often agree with PSS's vote, making em the most influential and suggesting that e had some magical way of influencing the outcome of decisions, or it could just be that e agrees with the populace most of the time. Although e has a relatively high average voting, G. has as significantly higher average voting strength, but a significantly lower result agreement rate. So it appears that this just indicates that e often agrees with the majority or that the majority often agrees with em (based on eir office-holding or vote comments), of which only the latter would make em an influencer. The most often that any voter disagreed with the result of a decision was 26% by L (mostly a zombie), and the most of any common voter was 23% by Alexis. Cuddlebeam's disagreement with other Agorans show up here, too. E disagreed with the result of decisions 22% of the time. **MOST INFLUENTIAL (perhaps): PSS** ## 3. The Most Influential Agoran Influencer summary: * Submitted proposals: Aris * Adopted proposals: Aris * Adopted proposal rate: PSS. * Adopted proposal words: Aris * Authored proposal margin (all): PSS * Authored proposal margin (adopted): PSS * Authored proposal margin (rejected): G. * Co-authored proposals: G. * Vote count: Jason * Voting strength max: Trigon * Voting strength average: G. * Endorsements: Jason * Endorsements (non-total) (tied): G., Jason * Determinative votes: Jason * Result agreement: PSS | Person | Influence Scores | Total | | :----- | :--------------- | ----: | | Aris | 1 + 1 + 1 | 3.0 | | PSS | 1 + 1 + 1 + 1 | 4.0 | | G. | 1 + 1 + 1 + 0.5 | 3.5 | | Jason | 1 + 1 + 0.5 + 1 | 3.5 | | Trigon | 1 | 1.0 | Based on this analysis, Publius Scribonius Scholasticus is the most influential Agoran and should therefore immediately be banished and reprimanded for daring to influence the game. Once that is done, Agora will finally be safe from the influencer and be able to proceed as a more equitable and less influenced society. ## 4. Analysis but without the pretense Now that the question burning on everyone's minds has been answered, there are a few other interesting datasets that I have available but don't fit in nicely with the influencer theme. ### 4.1 Voter-Voter Agreement By associating each vote resolution with a number (-1 for AGAINST, 0 for PRESENT, and 1 for FOR), the correlation between people's voting can be calculated. This yields the following chart:  Note that the agreement is only measured based on decisions where both voters voted. This potentially skews the results if people abstain when they don't have a strong opinion on a decision (and would be PRESENT if voting), but there's not a sane way to handle this with the data that I have. I have no way to distinguish between a voter intentionally or unintentionally abstaining, and I have no easy way to determine if a voter was even a player for any particular decision. This means that only counting decisions where both voters voted is the most sanely implementable way to measure this, even if it potentially results in skew. Looking at the top left of the chart, we see a sea of blue, indicating that the people who vote the most tend to agree with each other a lot. One small exception to this trend is Bernie & PSS, who happen to have a negative correlation. Bernie was a zombie for all of eir votes in these data, so this doesn't really indicate anything about Bernie emself, but eir masters over the time period -- Jason and twg, though both of them generally agreed with PSS overall, suggesting this is just an interesting anomaly. The next deviation from the trend is Cuddlebeam, who apparently disagrees with many of the common voters, and doesn't actively agree with the rest. This is sufficiently strong to suggest a pattern of Cuddlebeam being a _**rebel**_ when it comes to voting. Overall, however, this chart suggests that Agorans agree with each other a lot when voting. ### 4.2 Vote Kinds by Voter  Looking at the above graph, it is clear that FOR is by far the most common vote. This also suggests that Agorans agree with each a lot, but this time it suggests voters tend to agree with authors. One voter, however, takes this philosophy to the extreme -- Cuddlebeam. Cuddlebeam regularly votes FOR all proposals in a distribution while claiming not to have read them. In fact, e has only voted AGAINST a single proposal (out of the 69 e voted in) in these data: Proposal 8271. This likely contributes to eir low agreement rate with other Agorans. Other Agorans vote FOR a lot, but nowhere near as much as Cuddlebeam does. ### 4.3 Voter-Author Agreement By calculating the same vote numbers as above (1 for FOR, etc.) and averaging them for each voter's votes on each author's proposals, the agreement between the voter and the author can be calculated.  [Note: the number of votes by the voter decreases top-to-bottom, and the number of proposals submitted by the author decreases left-to-right.] Some interesting things are immediately apparent on this graph. One is Cuddlebeam's very dark blue line as a voter, showing eir pattern of voting FOR on most proposals. Another is the sea of blue to the top left, showing that the most common authors tend to receive many FOR votes on their proposals, as expected by their high proposal adoption rates. Yet another is D. Margaux's author line having both very dark blue and very dark red. The extreme values are expected, as e has only submitted two proposals. One of these proposals was controversial: ``` PROPOSAL 8209 (AFK Reform Act v1.1) AUTHOR: D. Margaux FOR (5): D. Margaux, Halian, L, R. Lee, nix AGAINST (9): Aris, Falsifian, G.$, Jacob Arduino, Jason, Murphy, Telnaior, Trigon, twg@ PRESENT (0): BALLOTS: 14 AI (F/A): 15/27 (AI=2.0) POPULARITY: -0.286 OUTCOME: REJECTED ``` while the other was not: ``` PROPOSAL 8307 (Deregistration) AUTHOR: D. Margaux FOR (0): AGAINST (10): Aris, Bernie, Falsifian, G.$, Gaelan, Jason, Rance, o, omd, twg PRESENT (1): Alexis BALLOTS: 11 AI (F/A): 0/31 (AI=3.0) POPULARITY: -0.909 OUTCOME: REJECTED ``` The people who voted FOR on Proposal 8209 did not vote on Proposal 8307, meaning they have a value of +1 for agreement with D. Margaux. Other than that, there are common speckles of red in a field of blue, indicating overall agreement rates of voters with authors are relatively high, confirming the earlier observation that the most common vote is FOR. ### 4.4 Endorsements by Endorsee/Endorser The below graphs show how often endorsements are by endorsee/endorser pair. Non-total endorsements are defined as above. #### All endorsements  An interesting thing about this graph is that it shows master/zombie relationships very clearly, since they are generally the darkest squares in the grid. For instance, it is clear that twg was G.'s zombie at one point and that Jacob Arduino was twg's zombie at one point. The next graph actually makes normal endorsements visibly distinct. #### Non-total endorsements  Here, it is clear who endorses other people the most - Falsifian. This is in part because e has frequently endorsed the author of a proposal in order to help allow the author to self-kill a proposal if e notices as bug in it. TWG has also endorsed people a lot, for what appears to be largely the same reason; in fact, e submitted a proposal that would have allowed authors to formally cancel proposals, so it makes sense that e tried to emulate that in voting. ## 5. Conclusion This thesis was based on a very silly premise, but hopefully you, dear reader, found it at least mildly interesting. The source code for the generation of the stats and graphs will be available in the AgoraNomic/assessor repository. ## 6. Acknowledgements * nix and G. for helping me develop the initial idea and letting me bounce stuff off them * everyone in the unofficial Discord server for tolerating the large amount of images and stats I posted, then reviewing the first draft of this thesis -- Jason Cobb Assessor, Rulekeepor, Stonemason
