<div dir="auto"><div>I just checked that Kemeny-Young makes it a tie between the finish orders ABC and BCA, while decloned Kemeny-Young puts BCA as the winning finish order ... ahead of ABC ... with BAC half way between ABC and BCA.</div><div dir="auto"><br></div><div dir="auto">It's interesting that if you sort BAC pairwise, you get BCA if you use Sequential Pairwise Elimination (starting at the bottom), while starting at the top (as in DMC) you get ABC as the sorted order. No wonder BAC turned out to be half way between ABC and BCA in decloned K-Y.</div><div dir="auto"><br></div><div dir="auto">And remember that BAC was the order preferred by the greatest majority over it polar opposite order CAB ... in other words BAC was the Swap Cost Approval winner.</div><div dir="auto"><br></div><div dir="auto">This suggests using Swap Cost Approval as the agenda setting method for SPE instead of Implicit Approval.</div><div dir="auto"><br></div><div dir="auto">-Forest<br><br><div class="gmail_quote" dir="auto"><div dir="ltr" class="gmail_attr">El lun., 21 de mar. de 2022 5:29 p. m., Forest Simmons <<a href="mailto:forest.simmons21@gmail.com">forest.simmons21@gmail.com</a>> escribió:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="auto"><div><br><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">El lun., 21 de mar. de 2022 2:13 p. m., Forest Simmons <<a href="mailto:forest.simmons21@gmail.com" target="_blank" rel="noreferrer">forest.simmons21@gmail.com</a>> escribió:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="auto"><div dir="auto">Here's an example of how to use River to get a finish order by making the potential finish order play the role of candidates:<div dir="auto"><br></div><div dir="auto">Let's use the following ballot profile:</div><div dir="auto"><br></div><div dir="auto">4 A>B>C</div><div dir="auto">3 B>C>A</div><div dir="auto">2 C>A>B</div><div dir="auto">1 C>B>A</div><div dir="auto"><br></div><div dir="auto">The respective faction favorite percentages are 40, 30, 30, while the respective anti-fovorite percentages are 40, 20, 40.</div><div dir="auto"><br></div><div dir="auto">The Swap Costs for converting the ballot order A>B>C into the respective potential finish orders </div><div dir="auto">ABC, BAC, ACB, BCA, CAB, & CBA, are ...</div><div dir="auto">0, 8, 12, 24, 28,& 36, so this order of increasing swap cost is the rank order of the potential finish orders induced by the ballot A>B>C. </div><div dir="auto"><br></div><div dir="auto">For example, the swap cost of converting ABC to CBA is the sum of the costs of transposing AB, AC, & BC, the cheapest sequence of swaps for converting ABC into its opposite order CBA. The sum of these elementary swap costs is given by ...</div><div dir="auto">[A\][B/]+[A\][C/]+[B\][C/]</div><div dir="auto">= (4×2+4×4+3×4)/100, or 36%.</div><div dir="auto"><br></div><div dir="auto">Similarly the costs for converting the ballot order B>C>A to the finish orders (..</div><div dir="auto">BCA,BAC,CBA,CAB,ABC,&ACB, are ...</div><div dir="auto">0, 12,12,24,24,&36.</div><div dir="auto"><br></div><div dir="auto">Costs of converting CAB to ...</div><div dir="auto">CAB, CBA,ACB, BCA,ABC,&BAC are</div><div dir="auto">0, 8, 12, 14, 18,&26</div><div dir="auto"><br></div><div dir="auto">Cost of converting C>B>A to ...</div><div dir="auto">CBA,BCA,CAB,BAC,ACB,ABC ...</div><div dir="auto">0, 6,12, 18, 24, 30</div><div dir="auto">Let's check CBA to ABC cost:</div><div dir="auto">[C\][B/]+[C\][A/]+[B\][A/]</div><div dir="auto">3×2+3×4+3×4 = 30</div><div dir="auto"><br></div><div dir="auto">Preferenc summary for finish orders ..</div><div dir="auto"><br></div><div dir="auto"><span style="font-family:sans-serif">40 ABC>BAC>ACB>BCA>CAB>CBA</span><br></div><div dir="auto"><font face="sans-serif">30 </font><span style="font-family:sans-serif">BCA>BAC=CBA>CAB=AB6C>ACB</span></div><div dir="auto"><span style="font-family:sans-serif">20 </span><span style="font-family:sans-serif">CAB>CBA>ACB>BCA>ABC>BAC</span></div><div dir="auto"><span style="font-family:sans-serif">10 </span><span style="font-family:sans-serif">CBA>BCA>CAB>BAC>ACB>ABC</span></div><div dir="auto"><span style="font-family:sans-serif"><br></span></div><div dir="auto"><font face="sans-serif">Pairwise scores ...</font></div><div dir="auto"><div dir="auto" style="font-family:sans-serif">BAC>ACB 80 to 20<br></div><div dir="auto" style="font-family:sans-serif">BCA>CBA 70 to 30<br></div><div dir="auto" style="font-family:sans-serif">ABC>ACB 70 to 30<br></div><div dir="auto" style="font-family:sans-serif">BAC>CAB 70 to 30<br></div></div></div></div></blockquote></div></div><div dir="auto"><br></div><div dir="auto">No other social order is preferred over its reversed/opposite order by such a large majority (70 to 30).</div><div dir="auto">This makes BAC a strong contender for "most acceptable" finish order, especially since the undefeated order BCA is only tied (50 to 50) with its polar opposite ACB.</div><div dir="auto"><br></div><div dir="auto">Anybody else think BAC should be considered the most generally acceptable?</div><div dir="auto"><br></div><div dir="auto"><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="auto"><div dir="auto"><div dir="auto"><div dir="auto" style="font-family:sans-serif"></div><div dir="auto" style="font-family:sans-serif"><div dir="auto">ACB>CAB 60 to 40</div></div><div dir="auto" style="font-family:sans-serif"><div dir="auto">ACB>BCA 60 to 40</div><div dir="auto">CAB>CBA 60 to 40<br></div><div dir="auto">BCA>BAC 60 to 40*</div><div dir="auto">BCA>ABC 60 to 40<br></div><div dir="auto">CBA>ACB 60 to 40<br></div><div dir="auto">ABC>BAC 60 to 40</div><div dir="auto">ACB>BCA 60 to 40</div><div dir="auto">CBA>ABC 60 to 40<br></div></div><div dir="auto" style="font-family:sans-serif">BAC>CBA 40 to 30<br></div><div dir="auto" style="font-family:sans-serif">ABC>CAB 40 to 30</div><div dir="auto" style="font-family:sans-serif"><br></div><div dir="auto" style="font-family:sans-serif">The River procedure applied to the above defeat strengths yields BCA as the winning alternative ... in fact, BCA is unbeaten pairwise, and tied only by ACB, the first candidate to be disqualified by River ... the strongest defeat being ACB's defeat by BAC, defeat strength 80 to 20.</div><div dir="auto" style="font-family:sans-serif"><br></div><div dir="auto" style="font-family:sans-serif">So BAC is the clear winner among potential finish orders.</div><div dir="auto" style="font-family:sans-serif"><br></div><div dir="auto" style="font-family:sans-serif">How about standard River? Both A>B and C>A are equally strong and compatible with the strongest defeat B>C. So there seems to be some ambiguity about which defeat to lock in next. If A>B is locked in next, then A>B>C would be the finish order. If C>A, is locked un next, then B>C>A would be the finish order.</div><div dir="auto" style="font-family:sans-serif"><br></div><div dir="auto" style="font-family:sans-serif">But we have just seen how applying Rver to the potential finish orders as candidates in their own rights (made possible by Swap Cost) totally resolves this ambiguity: BCA is unbeaten pairwise among potential finish orders.</div><div dir="auto" style="font-family:sans-serif"><br></div><div dir="auto" style="font-family:sans-serif">How about the standard (clone dependent) Kemeny-Young method finish order?</div><div dir="auto" style="font-family:sans-serif"><br></div><div dir="auto" style="font-family:sans-serif">-Forest</div><div dir="auto" style="font-family:sans-serif"><br></div></div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">El vie., 18 de mar. de 2022 9:41 p. m., Forest Simmons <<a href="mailto:forest.simmons21@gmail.com" rel="noreferrer noreferrer noreferrer" target="_blank">forest.simmons21@gmail.com</a>> escribió:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="auto">In an EM list message dated 13 Jan 2022 under the heading "Fixing Kemeny Young," I proposed a clone free cost function to replace the Kendall-tau metric that is responsible for the clone dependence of Kemeny-Young:<div dir="auto"><br><div dir="auto"><a href="http://lists.electorama.com/pipermail/election-methods-electorama.com/2022-January/003348.html" rel="noreferrer noreferrer noreferrer noreferrer noreferrer noreferrer noreferrer" target="_blank">http://lists.electorama.com/pipermail/election-methods-electorama.com/2022-January/003348.html</a><br></div><div dir="auto"><br></div><div dir="auto">Since then I have been referring to that function as "de-cloned Kendall-tau," but from now on, I will call it (more descriptively) "Swap Cost," because it is a measure of the general regret or disappointment resulting from the replacement of one candidate ranking with another ... and that cost is the sum of the costs of the transpositions or "swaps" that it takes to convert the original ranking into its replacement order.</div><div dir="auto"><br></div><div dir="auto">Click on the above link for a detailed description and rationale/explanation of the desired clone independence.</div><div dir="auto"><br></div><div dir="auto">The examples given hereafter will make the procedural details abundantly clear.</div><div dir="auto"><br></div><div dir="auto">For now, suffice it to say that in order to convert one ranking into another we need to know (for each candidate k) the fraction or percentage f(k) of the ballots that specify k as favorite, as well as the fraction or percentage f'(k) that specify candidate k to be "anti-favorite," or most disapproved.</div><div dir="auto"><br></div><div dir="auto">The disappointment cost of a single swap is the product f(k)f'(j) where, of two adjacent ranked candidates k and j, candidate k is lowered and j is raised, so that the order kj is transposed to jk.</div><div dir="auto"><br></div><div dir="auto">So the more k is top choice, the greater the disappointment from k getting lowered, and the more j is last choice, the more disappointment from j being raised.</div><div dir="auto"><br></div><div dir="auto">This directional sensitivity is totally absent in Kendall-tau for which every transposition incurs precisely the same unit cost, no matter which candidate is being raised or lowered.</div><div dir="auto"><br></div><div dir="auto"><span style="font-family:sans-serif">This directional cost dependence is like a taxicab fare where the time/distance from point X to point Y is not in general the same as the from Y to X.</span></div><div dir="auto"><br></div><div dir="auto">The classic treatise on this kind of path length asymmetry between "ida y vuelta" is a monograph entitled, "What is Distance?", by <span style="background-color:rgb(255,255,255);color:rgb(32,33,36);font-family:"google sans",roboto-medium,helveticaneue-medium,"helvetica neue",sans-serif-medium,arial,sans-serif;font-size:24px">Julij A. Šrejder.</span></div><div dir="auto"><br></div><div dir="auto">One of the greatest advantages of having a suitable cost function on rankings is that it allows a standard adaptation of any single winner Universal Domain method to be used as a basis for finding a finish order:</div><div dir="auto"><br></div><div dir="auto">The key is that just as a ranked ballot shows a preference of one candidate over another, the Swap Cost reveals support of a ballot for one finish order over another:</div><div dir="auto"><br></div><div dir="auto">If the swap cost of converting the ballot B ranking to finish order O1 is less than the cost of converting the B ranking to order O2, then we say that ballot B supports O1 over O2, or speaking anthropomorphically, "B prefers O1 over O2."</div><div dir="auto"><br></div><div dir="auto">Once we have a ballot based preference relation on the potential finish orders, we can use any Universal Domain single winner method to find that finish order.</div><div dir="auto"><br></div><div dir="auto">For example River. Unlike Ranked Pairs or BeatPath CSSD, River does not automatically produce a finish order in the course of finding the single winner.</div><div dir="auto"><br></div><div dir="auto">But once we have a preference relation (given by Swap Cost for example) on candidate rankings then we can apply the River formalism with the set of potential finish orders playing the role of the set of alternatives ... i.e. the potential finish orders become the "candidates," and the Swap Cost gives the (proximity inferred) ballot preferences among these candidate-alternatives.</div><div dir="auto"><br></div><div dir="auto">We'll do a detailed example of this tomorrow.</div><div dir="auto"><br></div><div dir="auto">But tonight let's just celebrate our new tool (Swap Cost) for converting any UD single winner method M (like River) into a finish order method similar to Kemeny-Young, but clone proof (unlike Kemeny-Young), provided the base method (like River) is clone-proof ... before Kristofer pops our bubble!</div><div dir="auto"><br></div><div dir="auto">-Forest</div><div dir="auto"><br></div></div></div>
</blockquote></div></div>
</blockquote></div></div></div>
</blockquote></div></div></div>