Impacts

The impacts of a decision-making tool may be felt by many types of stake holders. The tool may affect those who make decisions, those who participate in group decision-making processes, those who must live with the consequences of decisions, and those who interpret decisions. Here we assess the impacts of DSV from the perspective of three inter-related groups, policy-makers, voters, and analysts. The ability of voters (or survey respondents) to express themselves, policy-makers to make appropriate decisions, and analysts to interpret and explain election results will also impact a fourth group: those who must live with the consequences of the decisions.

Policy-Makers

In addition to its use as a voting system, DSV may be applied to group decision-making situations in which a single policy-maker must make a decision and may use a non-binding survey or the results of a previous election as a guide. In such situations, the impact of DSV on policy-makers can be analyzed by considering the accuracy with which DSV discovers voter preference information and the extent to which DSV facilitates arrival at a decision.

A known flaw of nearly every method for obtaining election or poll input concerns the extent to which the voters can, should, and do present insincere preference information. DSV, in particular the ballot-by-ballot method, is the first system that elicits sincere preference data without excessively relying on randomness. As compared with other systems, DSV provides clear advantages in terms of discovering sincere preference information from voters.

DSV also facilitates arrival at a decision by highlighting preference patterns that might be overlooked using other decision-making tools. Although further work would be useful for establishing guidelines policy-makers can use in interpreting DSV results, we believe DSV is useful to policy-makers in its current form. The following are examples of areas in which DSV can aid a policy-maker:

• DSV is useful for gauging strength of preference and level of indifference among voters. It can help distinguish alternatives that are strongly favored by their supporters from those that are only slightly favored by their supporters. For example, in our course times survey simulation, the use of DSV revealed that students who preferred 8:00 am classes preferred them only slightly to 6:00 pm classes, whereas students who preferred 6:00 pm classes strongly preferred them to classes held at any other time.
• DSV is useful for identifying sets of alternatives that act as mutual substitutes. These alternatives are all highly rated by the same group of voters, and these voters are either indifferent between members of this set or only weakly prefer one member of the set to another. For example, in our university lecture series survey simulation, DSV revealed that ``science'' and ``health and medicine'' were mutual substitutes, favored by the same group of respondents.
• DSV is useful for determining which alternatives voters would support if they knew their favorite was not a serious contender. This allows decision-makers to collect one set of ballots and use them to simulate a run-off election or survey.

Voters

We have also shown that DSV has properties that make it attractive from the perspective of voters. DSV is sufficiently general that it subsumes all other voter preference expression mechanisms currently in use. While such generality is interesting at a theoretical level, there is a far more compelling reason to claim superiority for DSV with regard to expression of voter preference: DSV is the only system to date that simultaneously allows voters to express sincere preference information while still casting their votes strategically. While it remains true that not every voter will be sufficiently educated to develop the best preference profile, DSV affords each voter identical access to the preferences expressed by others. Moreover, the rationally optimal strategy formulator described in Chapter  affords all voters equal ability to vote strategically based on their preference profiles.

The availability of computers that can perform relatively sophisticated computations quickly enables us to implement DSV systems in which the burden of strategy formulation does not fall on the voters. In addition, computers make possible systems in which many rounds of balloting may be conducted quickly and without returning to the voters. As a result, voters are able to express themselves sincerely without diminishing the effectiveness of their votes. Furthermore, although computers are necessary for tallying ballots under DSV, they are not required for ballot collection: voters can register their preferences for each candidate on a paper ballot rather than entering them directly into a computer. Thus, DSV offers the following benefits to voters:

• DSV allows voters to vote strategically (thus maximizing the effectiveness of their votes) regardless of whether they have obtained accurate information about the preferences of the rest of the voters or whether they know how to formulate optimal strategies.
• DSV allows voters to express their sincere preferences (including their preference intensities) without sacrificing the effectiveness of their votes.
• DSV allows voters to express their preferences for all contingencies, allowing multi-round elections to be conducted with a single round of balloting.

One drawback to ballot-by-ballot DSV in its current form is that it is unstable, and thus election results may not be repeatable and may be dependent on chance. Although repeating the election many times and averaging the results may prove satisfactory in some circumstances, our acceptability study suggests that voters may have trouble accepting results based in part on randomness.

Analysts

DSV is also useful to those who interpret election results, including scholars, media analysts, and elected representatives who want to understand the opinions of their constituents. To evaluate the impact of DSV on analysts, we examine the kind of information DSV makes available to the analysts and how such information contributes to the analysis.

To the extent that DSV elicits sincere preference information, it is clear that DSV provides more information than do traditional voting systems, in which analysts cannot distinguish sincere from insincere votes. There is still much to be discovered concerning how such information can be useful in explaining voter behavior or understanding the results of a DSV election. In addition to any final tally reported for a DSV election, we believe that analysis of intermediate results may be crucial to obtain an accurate view of election results. Consider, for example, 1000 simulations of a ballot-by-ballot election. At the end, candidates A and B may each show an average of 50% support, but this could occur under the following disparate situations:

• In each round, candidates A and B received approximately half the voters' support.
• In each round, either candidate A or B received nearly all the support, but half of the rounds elected A and the other half elected B.

Voter behavior is quite different in the above two situations, yet this cannot be seen merely by examining the end-result of DSV computations. Further research is needed to determine how the information harvested from a DSV election can help explain voter and election behavior. It is our belief that while DSV makes analysis more difficult, it will ultimately make analysis more revealing.