WO1998012666A2 - Method, machine and system for counting audiences - Google Patents

Abstract

Disclosed is a machine for counting the members of an audience. The machine registers responses from the audience through a novel sampling method and multiplies the number of responses by a stored multiplier to yield an approximate audience count. The machine is enhanced by features for adjusting the multiplier to audience characteristics. Also disclosed is a decision system for use when parties disagree on which alternative to pick. The system registers the votes for each alternative and assigns each alternative a probability of being picked corresponding to the number of votes it has gotten. The system then picks an alternative by random number generation. The system can be used by single judges to decide upon alternatives by assigning each alternative a probability of being picked. Also disclosed is a system for authenticating financial messages.

Description

Method, Machine and System for Counting Audiences

Introduction

A company placing an ad wants to know how many people see or hear the ad. A company also wants to know who is using its product or service. Both problems are counting problems, how to count audiences (users can be considered a kind of audience, for they see or hear the product or service they are using).

Here we present a novel method for counting audiences. The method can be implemented in a machine, and the machine can be embedded in a larger system for connecting audiences to the machine. Thus we say that the invention we are describing is a method, a machine and a system. Wc explain the basics of the method first and then describe the machine and system.

In addition to counting an audience, a company wants to know about the audience, who the members are — their age, sex, zip code, income and so on. Even more, a company wants to know whether an ad is effective, whether a product is used, how it is used.

The inventive method (IM) is not only a method for counting but also a method for obtaining a random sample of an audience. By querying this sample it may be possible to find out about the behavior of the audience.

Summary of the Invention

The invention is a machine for counting the members of an audience. The machine registers responses from the audience through the IM and multiplies the number of responses by a stored multiplier to yield an approximate audience count. The machine is enhanced by features for adjusting the multiplier to audience characteristics.

The Goal of the Invention

The goal of the invention is to enable a organization to find out:

1) how many people have seen or heard its ad, product or service,

2) characteristics about those people, such as their age and zip code,

3) the behavior of tho.se people with regard to the ad, product or service (e.g., whether the ad has been effective). While it is obvious that advertisers need "ratings" to make intelligent advertising decisions, the same need exists for providers of products and services. The better one can count users, the better one's economic decisions, in general. For example, a park designer who puts a tennis court in a park will want to know how many people use the court in order to decide whether to build another.

Scope of the Invention

The IM can be used to count audiences of radio, TV, print publications (including photocopied material), websites, billboards, direct mail pieces, fliers, and so on. Likewise, it can be used to count users of most goods and services. In other words, the IM is a general way to count the set of people who see, hear or use virtually anything.

Note on Terminology

For convenience in the discussion below, we will use the term ad to refer not only to an advertisement but anything, tangible or intangible, that has an audience that is to be counted. Using the term ad gives some concreteness to the discussion and has the added benefit of being short. We will use the term company to refer to any organization or person who wants to count and, possibly, analyze an audience.

The Basic IM

The basic idea behind the IM is to add a match message (MM) to an ad. The MM would match a known percentage of the ad's audience. Further, each person who matched the MM would be rewarded for responding to a counting entity in order to be counted. A person who matches an MM will be called a matcher.

The MM would normally be a universal type of identifying number that virtually everyone in an audience has in common, such as a birthday or social security number (or part of any such number). The key to the IM is that the percentage of people who share a given instance of the universal number, say a given birthday, is roughly known. For example, if January 1 is the MM, it would be known that roughly 1 in 365.25 people share the number. Or if, say, the last three digits, 333, of a social security number were the MM, it would be known that roughly 1 in 1000 people have a social security number ending in that number. As for the reward for responding to a match, it might go like this: "Anyone who sees this ad and was born on January 1, call 1-800-COUNTER and collect $50."

If the percentage of people who match the MM is known then the total audience count can be arrived at by counting the people who respond to the offer. That is the basic idea behind the IM.

Note, the people who respond are, in effect, a random sample of the ad's audience. The MM can be chosen so that the sample is not random, but is skewed to some characteristic, such as choosing women over the age of 55, but that is not the point here. The point is that the response to the MM creates a desired sample of the ad's audience. Extrapolating from the sample gives a total count of the people who the sample represents.

Of course, many complications stand in the way of an accurate count. We will address these later. For now, the basic idea is given. To repeat:

1. To an ad, attach to an MM (an identifying number or other identifying data) that matches some known percentage of the ad's audience.

2. Offer people who match the MM a reward for responding to a counting entity (e.g., calling an interactive voice response system that registers responses).

Note: The message stating the reward could be standardized by some kind of trademark. For example, if a person sees a number inside a certain symbol then she knows that there is a reward for calling a given toll-free number, if she genuinely matches the MM.

Also note: since the IM will be used to count audiences for different ads, each ad needs to be identified somehow. An ad might be identified by a number, to distinguish it from other ads, or it might just be described, e.g., "The McDonalds ad at half-time on the Superbowl," or "The McDonalds ad on page 3 of the LA Times, August 1 , 1997." It is also possible for the MM to identify an ad because different ads will usually include different MM's.

3. Calculate from the number responders what the total audience is. About Matching Messages

Some useful MM's include birthdays, government ID numbers and telephone numbers. An MM does not have to be universal, as long as the percentage of the population that has the MM is roughly known. For example, not everyone has a phone number in their name but the percentage of the population that does is roughly known.

A base MM can be modified by additional data to make a more specific MM with a different matching frequency. For example, a birthday can be modified by age and sex specifiers. For example, the MM can be: All men born on January J, before 1960.

Benefits of the IM

1. The IM is often cheaper than current audience counting methods.

2. In many situations, the IM is more precise than current auditing and sampling methods. It enables a company to check the audience for very specific things. It enables a company to overcome the generic problem of using one audience to represent another. For example, a company can check the audience for a given ad or article in a magazine, rather than extrapolate from the circulation of the magazine. As another example, it enables a company to avoid channel surfing type problems and find out how many people are seeing an ad not just a TV show.

3. The IM enables a company to count sub-audiences. For example, a company can check how many women over 55 are using a product.

4. The IM enables the efficient counting of all kinds of audiences not previously practical to count. For example, it enables a company to find out how many people read a photocopy of an article.

5. Where usage can be registered through automatic means, as for example counting how many people visit a website, the users are usually anonymous. The IM enables a company to find out about the audience without violating the privacy of the total audience. Simplest IM Counting Machine

A special purpose counting machine (CM) can based on the IM. The CM is a computer (with means for entering, storing, manipulating and outputting data) that includes functions that implement the IM. We will first describe the simplest such CM and then later describe additions that make the CM more accurate.

In order to arrive at an audience count for an ad, the CM requires means for registering:

1. data that identifies the ad whose audience is being counted,

2. the MM that is associated with the ad.

The CM will also have to store a multiplier associated with the MM.

The CM will also have means for storing a responder's ID data. (Although this information is not essential for arriving at a count; it enables the CM to authenticate a responder and to pay a reward.)

The CM will execute the following steps when it receives a response:

-register the ad ID data,

—register the MM,

—add to the tally of responses for that ad,

-pull the MM multiplier (the multiplier associated with that MM),

-multiply the tally by the multiplier to yield an audience count,

—output the count.

(As noted, the CM will also normally register the responder's ID data so that the responder can be paid. Normally, in fact, more data than name and address will be gathered. Here, though, we are just describing a minimal CM that implements the LM.)

Let us now give an example. Our audience will be the people attending a football game. The audience will be reached with an ad shown on the scoreboard. The scoreboard shows the following ad and MM:

"Pete's Wicked is better than Bud. If your social security number ends in 333, call 1-800- COUNTER when you get home and get a $50 reward." Now, when a responder calls in, an operator asks her what the MM is and takes her ID data. We'll assume that the MM identifies the ad in this case, thus "333" tells the CM which audience is to be counted.

The CM pulls the multiplier for "333", which we'll assume is 1000 (we're assuming the last three digits of a social security number are random).

The CM tallies the responses for the MM "333". Let's assume that 50 people respond. The CM multiplies this tally by the multiplier yielding an audience count of 50,000.

Placing the Matching Message

In certain cases, such as with website applications, it is possible for the CM to include means for attaching an MM to an ad. In such cases, the CM is a closed counting system which originates an MM and collects the responses. In most cases, though, the MM will be placed by humans. In any case, the CM must reflect the placement of the MM in an ad — the ad must be associated with the MM within the CM's memory (database).

The CM's Interfaces With the Audience

The audience can interface with the CM in three basic ways. The CM can include means for receiving data in all three ways.

1. Human Operator Interface

A responder can call in to an operator who takes data and enters it into the CM. At minimum, the operator would take the following information: a) the responders' ID data, b) the MM, c) the ad's LD data.

2. Interactive Voice Response Interface

The CM can include interactive voice response (IVR) means and thus receive data directly from responders.

3. Web (Online) Form Interface

The CM can include a web interface for receiving data directly from responders. A web interface is especially natural where the IM is being used to track web usage. In this case, a responder can click on a hypertext link that takes her to a CM website where she can fill out a CM form. In this case, the minimum data that needs to be filled in is only the ID data of the responder (because the rest of the minimum data can be captured from the originating website). Of course, the form can include extra fields for entering any kind of information that the CM is designed to capture.

Gathering Additional Data About the Responders

An advantage of the IM is that responders can be asked various questions that can reveal additional data about the audience they represent. Depending on the questions, this data can indicate not only who is in an audience but whether an ad has achieved its purpose (we will take this topic up further in the section on evaluating the effects of ads).

At minimum, the CM can include means for registering minimal demographic data from responders, including sex, age, income and postal code.

The CM can capture such data through an IVR or an online form. If the CM includes means for capturing such data then it is more than a counting machine; it is a system for gathering data about an audience.

Analyzing Responders

The CM can also include means for analyzing the additional data that it gathers. The CM can include means for sorting responders according to demographic data, and can thus output the number of people in the audience who fall into various demographic categories — for example, how many people in the audience are women from 50-60 years of age and who live in the 2001 1 zip code.

Complications

The IM will yield an accurate count if:

• we know exactly what percentage of an audience matches the MM,

• we get a 100% response from matchers and,

• there are no cheaters.

Of course, these assumptions are unrealistic. Thus, for the IM to give an accurate count, the CM needs to include means for:

• correcting for the match characteristics of audiences,

• correcting for a less than 100% response rate,

• correcting for cheating and

• preventing cheating.

We will take these issues one at a time, starting with means for preventing cheating.

In this section we will also introduce the term managers to refer to the people who run the CM.

Basic Cheats Against the IM

Since the IM works through rewarding responders, there will be people who want to falsely collect rewards. These people need to be deterred because they corrupt the results of the IM. There are at least a few basic cheats that are hard to block by direct observation.

One is where a cheater uses the identity of a confederate. For example, if Joe sees an MM of January 1 and he knows that his brother's birthday is January 1 , he can claim the reward and falsely authenticate his identity by, say, sending in a copy of his brother's driver's license.

Another cheat is where a person tells a confederate that she matches an MM that she has not personally seen or heard. For example, if Joe sees an MM of January 1 , and knows that his friend, Jane, was bom on January 1 , he can tell her to call in for the MM reward, even though she did not see the MM herself.

Another fundamental cheat is "overviewing", where a person will simply scan ads just for their MM's. (This can even be done in an automated manner, in some cases.)

Detecting and Disqualifying Outliers

A basic way to stop cheaters is to rely on the fact that they will be extreme outliers. We say "extreme outliers" because in any normal distribution there will be outliers. We are referring here to extremely improbable response rates. We'll use the term outlier from now on to mean extreme outliers. By detecting outliers, the CM can detect cheaters. To detect outliers, the CM needs to keep a record of all the responses for each responder. The response rate for a given person can be subjected to various statistical tests to determine whether the person is an outlier.

The simplest test is to compare a person's response rate to that of everyone else's — in other words to place the responder in a distribution.

Other tests can take into account the different odds involved with each response. The probability of each match that a person claims is, obviously, key to determining whether a person is an outlier. For example, a person may respond to an MM that 1 out of a 100,000 people match and respond to another MM that 1 out of 100 people match. And so, the CM can include means for taking into account the probabilities (the multipliers) associated with MM's.

The point here is not to specify the statistical tests for determining whether a response rate is abnormally high; the point is simply that the CM can include means for doing statistical test to detect outliers.

To detect outliers, the CM executes the following steps: —registers a response, —registers the responder's ID data, if there is no file for the responder, creates a file and stores the response data, including MM multiplier, in the file, if there is a file, stores the response data, including MM multiplier, in the file and performs statistical tests on the file to determine if the responder is an outlier, if the responder is not an outlier, does nothing, if the responder is an outlier, outputs a flag.

Managers can institute policies that make it tough for cheaters to profit from their efforts. One such policy can be that rewards are withheld for a year until it is determined whether a responder is an outlier. Another policy can be to make rewards high enough to get a reasonable response rates but low enough so there is little incentive to cheat.

Keeping Track of MM Placement

Managers need to disperse MM's so that the chances that the same person will sec a particular MM are low. The chance cannot be set precisely because managers don't know the exact habits of individuals. Still, the idea is to not unnaturally expose a person to MM's that she matches. To take an absurd example, if a newspaper was to include the same MM on it's front page every day then it would be quite reasonable for a set of people to make numerous responses. They would be honest and they would be outliers. To avoid situations like this one, and to make cheating harder, it is important for managers to keep track of the placement of MM's in ads.

Thus, along with means for recording a non-descriptive ad ID number, the CM can include means for recording when and where each MM has been placed. Such records enable managers to decide which MM's to place with which ads.

Timing Measures for Certain Media

For certain media, especially TV and radio, one way to stop cheating is to have a short window of time for responders to contact the CM. Thus, the CM can include means for timing out the eligibility to respond.

For example, an MM can be shown during a break in a TV show with the proviso that a matcher is eligible to collect the reward only if she calls within 10 minutes of the MM being shown. The CM has the time deadline in its records and can disqualify the responder.

Thus, the CM can include the following steps: —register a time deadline for an ad, —register a response to the MM for the ad, —check if time deadline is passed, if yes, output a disqualification flag.

Interviewing Responders

Another method for correcting for cheating is to get data on the cheating rate. Responders can be asked if they cheated in any way. (Of course, the the responders must be are encouraged to be honest.) By gathering empirical data, managers can apply correcting factors (CF's) to MM multipliers. The CF's will differ depending on demographic data since people in different demographic categories will have different propensities to cheat.

A CF can be a coefficient or, perhaps, a mathematical function.

Using CF's, based upon empirical data, is also the main way to correct for response rate variability, a subject we take up next. Correcting for Response Rate Variability

In addition to the possibility of cheating, another problem with the IM is that the rate of response to MM rewards is unpredictable. 100% of the people who see an MM will not respond to the reward offered. Response rates will vary depending on the amount of the reward and on the demographics of the matchers.

A related problem is that 100% of matchers will not see the MM, even though they see or hear the ad that the MM is attached to.

The solution to these problems is to gather data about response rates and then convert this data into CF's.

Data can be gathered in various ways. For example, the MM reward can be varied from $1 all the way up to, say, $500 (an amount that nearly 100% of people would respond to). That way. the response levels per dollar reward can be charted. A table can be generated that lists response rates that correspond to pairings of demographic categories and reward amounts. The table can be included in the CM.

Managers may also need to gather data for situations they think are relevant. By situation we mean a set of circumstances that surround the MM in an ad. The kind of ad might be an important factor. For example, a TV ad might have a different CF than a print ad. The placement of the ad might be another. For example, a TV ad at 7:00 AM on Saturday on a children's' show might have a different CF than an ad at 7:00 PM on Saturday on a newsmagazine. For each relevant situation, the CM would include a CF. When a response is entered, the relevant situation identifier is also entered, and the CM pulls the corresponding CF. Of course, more than one situation might be involved.

To correct for variations in response rates, the CM executes the following steps:

—stores a table of CF's that correspond to pairings of demographic categories and reward amounts,

—registers a response to an ad,

—registers ID data for the responder,

—registers the MM for the ad,

—registers the reward amount for the MM,

—registers the demographic category for the responder, —pulls the CF that corresponds to the pairing of the reward amount and demographic category registered,

—applies the CF to the MM multiplier for that response, —multiplies the corrected MM multiplier by the response (which is 1 ), -adds the corrected figure (MM multiplier x 1) to the audience tally.

(Here we are describing how each response — each matcher who responds — is given an MM multiplier. Whether the CM stores multipliers for audiences depends, of course, on what tables and functions the CM includes, and on how responses to MM rewards differ among audiences. We assume that the MM multiplier depends on demographic data but it does not have to. It could just depend on the reward level, for example.)

Correcting for Variations in the Match Frequencies of Audiences

Ideally managers would know exactly what percentage of an audience matches an MM. In reality, this frequency will not be known precisely.

Data can be gathered about match frequencies for particular audiences. For example, the percentage of people whose birthday is January 1 might be different in different parts of a country. Random samples can be drawn from the relevant population, say, the population of Minneapolis, to determine what percentage of people have each birthday, Jan. 1 - Dec. 31. As another example, the percentage of people who have a social security number can vary among the audiences of different magazines.

To correct for variations in match frequencies among audiences, the CM would include an audience/MM multiplier that the corresponds to the pairing of a given MM and audience category (audiences would have to be described in some way, of course).

An audience MM multiplier is based upon empirical data rather than just assumptions stemming from the nature of the MM. For example, if the last three digits of a social security number are used as an MM, a natural assumption is that one out of every one thousand people share the same last three digits. This assumption is not necessarily correct. The actual frequency will depend upon the audience. And so, match frequencie — MM multiplier's that is — can be tailored for particular audiences. In order for a CM to use audience/MM multipliers it executes the following steps: —stores a multiplier to correspond to the pairing of an audience category and MM, —registers a response to an ad,

—registers the audience category corresponding to the ad, —registers the MM for the ad,

—pulls the multiplier that corresponds to the pairing of that audience and MM, —multiplies the response by the multiplier and adds the result to the audience count.

(Note: rather than storing audience/MM multipliers, the CM can store CF's that correspond to audience, MM pairs. The CF's would be empirically derived. Such a CF would be applied to a base MM multiplier which would be assumed. For example, lOOOx might be the assumed multiplier for an MM that is the last three digits of a social security number. The audience MM CF would be multiplied against the base multiplier for that MM. There is no substantive difference between this method and the one given above.)

Finding Out Whether an Ad Is Effective

The holy grail of advertising is to be able to find out the effect of an ad on an audience. Two fundamental obstacles stand in the way of evaluating the effects of ads.

One is that it is hard to observe the effect of an ad. Where direct response advertising is concerned, there is at least a response that is observed and linked to a specific ad (although the linkage is not certain). But with most kinds of ads, there is no directly observable response at all. The second obstacle is that an ad is part of a multi-factoral situation. If an ad is designed to sell a product, it is hard to say that the ad alone sold the product. Many other factors, such as a company's reputation, come into play. It is hard to disentangle the contribution of a given ad.

How can the IM change these fundamental problems? Well, the IM cannot change the problem of multi-factoral situations. Only good experiments, in certain situations, may be able to do that.

What the IM can change is the ability to observe. It enables a company to inexpensively get a random sample of an ad's audience. Previously it was very difficult to isolate who was exposed to an ad and when. Thus the IM makes it practical for a company to observe an ad's audience. (The observations won't be complete, of course. A company won't be following people around with cameras, and it won't be inside people's minds, but it will be able to identify a representative sample of people exposed to the ad, and the approximate time they have been exposed.) With a random sample of an ad's audience, a company can do a poll, it can compare the effect of one ad to another and, it can set up a controlled experiment. For example, a company might ask the sample:

1. Had you heard about our product before you saw this ad?

2. Did you buy our product before you saw this ad?

3. Did you buy our product after you saw this ad?

As another example, the headlines of two ads can be compared. A split run test can be done where everything about two ads is kept the same except for the headline. The IM samples for each ad can then be queried to find out which headline was better. (In fact, the IM samples may be significantly different due to the headlines.)

Another way IM samples can be used to check the effect of an ad is through the use of traced coupons. Where coupons are offered simply to test an ad's effect, the IM enables a cheaper way to do the test. Thai's because coupons can be given only to the IM sample, which is much cheaper than giving coupons to the full ad audience.

As mentioned, IM samples can also be used to do controlled experiments to measure the effect of ads. The ideal of a controlled experiment to measure the effect of an ad is to observe two groups of people that are exactly the same except that one group has been exposed to the ad and the other has not. One group can be an IM sample. The behavior of the IM sample — e.g., how many people in the sample bought the advertised product — can be compared to a control sample, a random sample of the population. But, as we'll see in the example below, a random sample might not suffice as a control. A skewed sample might be needed. And it might be necessary to draw it by the IM as well.

Finding a control sample is not so easy. A purely random sample drawn from the general population usually will not do because the idea is that everything about the two groups is supposed to be the same except the ad exposure. The very fact that a group of people would be exposed to an ad makes that group different from a random group.

For example, if the ad to be tested is a magazine ad for Rogaine, a baldness cure, then the group to be observed is the group that sees the Rogaine ad. Thus, the comparison sample, the control, needs to be of a group of people who would either stop to look at a Rogaine ad (or stop to check the MM in the Rogaine ad). The propensity of the control group to stop for either reason should be the same as for the group that actually is exposed to the ad. A random sample will not be a good comparison sample (control group) because the random sample's propensities will not be the same as the propensities of the group that actually stops to look at the Rogaine ad. One way to draw a comparison sample with roughly the same propensities is to include an MM in a "partial ad" with a headline the calls out the area of interest of the Rogaine ad. The headline might say, "Baldness Cure," and nothing else. The ad itself would otherwise be blank but be of the same size as the Rogaine. An IM sample from this audience will not have seen the Rogaine ad but should share all the other factors of the audience that did see the Rogaine ad (presuming a split run test is done so that ad placement and general demographics do not differ from those of the test ad).

(As noted, the term "ad" in this specification has been used to represent not only advertisements but anything that has an audience. In the discussion about evaluating ads, the focus has been on advertisements themselves but, the ideas extend to other things. It is a question of what reactions are being measured. For example, if a drug company wants to test the effectiveness of its acne medicine on buyers, it can take an IM sample and compare the reaction of that sample to the reaction of a control sample. Advertising was emphasized because the IM gives advertisers in particular a new tool.)

The IM as a Way to Do Relative Ratings

A count that the IM yields may be inaccurate. It may be skewed. Still, as long as the skew is reasonably consistent per category of ad, say, TV shows on Wednesday at 8:00 PM on ABC, the IM can be valuable as a source of relative ratings. A similar comment can be made about current ratings systems. They may or may not give accurate counts but at least they give a reasonable means for comparison. Of course there are controversies with current ratings systems because they seem to favor certain kinds of shows. The same may or may not be said of the LM. The IM may be more accurate as a relative measure than current ratings systems. Why? Because, it seems possible to gather good empirical data about the response rates of different demographic groups to IM rewards. Once the response rates per demographic group are found out, it should be easy to correct for the results of an IM survey. For example, if 50 people respond to an IM reward, it should be possible to break this sample down into the individual people who have responded and sort them demographically. Then a separate multiplier can be applied to each person to yield the total count. Machines As Audiences

An audience being counted by the IM does not have to be made up of humans. It needs to be made up of entities that an MM can be applied to and that have the ability to respond to an MM. For example, a numbered computer, such as a computer with an IP number, could respond to a census done through the IM. A central computer could broadcast an MM to a set of computers on a network and those computers that matched the MM could respond. Of course, this scenario assumes infrastructure and software that is not in fully in place. Further, the scenario assumes that a sampling method is desirable for taking a census of machines.

Random Number Method and System For Making Fair Decisions

Field of the Invention

The invention relates to a method and system for making decisions.

Discussion of Prior Art

The random number generator (RNG) is one of humankind's great inventions. One area where RNG's are used is to help decide among alternatives. For example, if a dozen people are vying for one apartment, the person who gets it might be determined by lottery. As another example, if two friends want to go to dinner together but disagree on the restaurant, they might flip a coin to decide where to go.

This patent application discloses a new method using an RNG for making decisions. The method can be used in judging and voting.

Let us take the case of judging first. As an illustration, assume that a judge in a divorce case has to decide who gets custody of the couple's poodle. The judge can say, "I think the husband is 40% right and wife is 60% right." One way to decide who gets the poodle is to assign husband a 40% chance of winning and the wife a 60% chance and then let an RNG do the deciding.

The case of voting is similar but a little different. In this case, people vote for the alternatives in question. The vote tally per alternative then determines a probability weight for each alternative. And then an RNG decides which alternative is picked. For example, say a group of five people have to decide between an Italian and a Chinese restaurant. And say that the Chinese restaurant gets 4 votes and the Italian gets 1. Well, a number could be randomly selected from 1-5. If it turns up 1-4 then the group would go to the Chinese restaurant. If it turns up 5 then the group would go to the Italian restaurant.

Has the method been used before? The answer is, yes, in certain simple situations. For example, it was used in the flip-a-coin-to-pick-a-restaurant example above. But, it does not appear that people have realized the generality of the method. So, as far as the inventor knows, the full method and its uses have not been disclosed. Summary

The invention is a method and system for judging and voting that uses a random number generator to choose an alternative from a set of two or more alternatives. In the case of judging, a judge considers the alternatives at hand and assigns each a probability weighting based on the merits of each alternative. Then a random number generator picks one of the alternatives. In the case of voting, people cast votes for the alternatives at hand. Each alternative is assigned a probability weighting corresponding to the number of votes cast for the alternative. Then a random number generator picks one of the alternatives.

Brief Description of Drawings

There are no drawings. They are not necessary to illustrate the invention.

Description of the Invention

Consider a hypothetical court case. Jack has slipped on Jill's icy walk and is suing Jill. Jill says that she put cat litter on the walk and further posted signs warning people that the walk was slippery. Now the judge, Wapner, has to decide who wins the case. The judge thinks to himself, "Jack is around 40% responsible because he was an oaf who did not pay attention to signs. But Jill is around 60% responsible because she should have made her walk safe for all visitors. The law is not clear here. I guess I'll have to pick Jack." So Jack wins. Now, if we take all the similar cases where the Jill's are "60% responsible" we find that the Jacks win 100% of the time. This situation seems unfair. Why not let the Jill's win 40% of the time?

Consider the passage of a tax law in a legislature. Lets' say the legislature makes laws by majority vote. And let's say that 60% of the legislature favors the tax and 40% does not. So the tax passes. That may or may not seem fair. But now let's say that there are two factions in this legislature. And let's pretend the members of each faction always vote together. And, finally, let's say that one faction makes up 60% of the legislature and the other 40%. Well, the 60% faction wins 100% of the time. This situation seems unfair. Why not let the minority faction win 40% of the time?

This application discloses a decision making method that utilizes a random number generator and that may be considered a fair way to make decisions in certain situations, including, perhaps, those above. A system for implementing this new method is also disclosed. For future reference, the name of the invention, Random Number Method and System for Fair Decisions, will be abbreviated as: Roll'em (its easier to say than RNM&SFFD and gives the impression of a random number generator in action).

The method can be used in most situations requiring judging and voting where the interests of parties conflict. However, the method is usually not appropriate. When is it appropriate? Well, the inventor is not sure.

Proper decision making is a mystery and the inventor is not sure in what situations Roll'em applies well. The invention will be described a general method and system for choosing between abstract alternatives. What these alternatives actually are is up to users.

Generally, when a good argument for property rights can be made by the parties involved then Roll'em may be appropriate. If we must make a general statement of how Roll'em can be used: it is a general way to dividing or spending common property.

Random Number Generator Defined

For our purposes, an RNG can refer to any of several means for supplying random numbers (RN's). It can refers to a device that generates pseudo-RN's from an algorithm. It can refer a device that uses RN's that are stored in it's memory. It can refer to a device that pulls RN's from an independent source of RN's. In other words, it is the functional part of a machine that is responsible for supplying the RN's used by the machine.

Roll'em for a Single Judge Making a Decision

Judges, at least in their minds, often use numbers to express the merits of the case of each claimant in a dispute — e.g., "John is 40% responsible and Sue is 60% responsible," or "John is 40% right and Sue is 60% right," or "John has 40% of the rights and Sue has 60% of the rights." When a judge assigns numbers like these we don't really know what they mean and she probably doesn't either.

But, as mentioned, in certain cases it is reasonable to convert these "merit" percentages into probability weightings that arc used to select the winning claimant by random number generation. A judge may be deciding from among more than two competing claimants, of course. The Roll'em method still applies — the judge simply assigns probability weights to the cases of all the alternative claimants.

As described below, roll'em can be implemented in a computer to enable a judge to decide among alternatives.

Note in the implementation below, even though a single judge is making a decision, we are going to say that probability weightings are made up of votes tallies. Here the idea is that the judge casts mental votes. The usage may be confused with conventional votes but let us think, then, of the judge having a group of voters in his mind, each casting a vote for the alternatives that the judge is weighing.

The invention comprises a computer including processing means, memory means, input/output means, and display means that executes the following steps:

1. Input information identifying the case being considered.

2. Input the alternatives being considered.

3. Input N, the total number of votes to be cast.

4. Input the votes cast for each alternative.

5. Register the number of votes cast (the probability weighting) for each alternative.

6. From the set integers, 1-N, assign to each alternative a set of winning integers, the set being equal to the number of votes registered to each alternative.

7. Get a random integer from 1-N.

8. Find the alternative that has an integer assigned to it that matches that random integer.

9. Output that alternative as the winner.

Roll'em for Voters Making a Decision

As discussed, Roll'em is also suited for voting situations. In this application, votes are cast for alternatives and the relative numbers of votes cast for each alternative determine the relative chances that each alternative has of being picked by a random number generation.

It should be noted that this application of roll'em works where voters have one vote each, as in the U.S. Congress, or where voters can have any number of votes, as in a shareholder proxy fight, where a shareholder's votes correspond to the number of shares she holds in the company.

As described below, roll'em can be implemented in a computer to enable voters to decide among alternatives.

The invention comprises a computer including processing means, memory means, input/output means, and display means that execute the following steps:

1. Input information identifying the issue being considered,

2. Input the alternatives being considered.

3. Input a voter's ID information.

4. Input the voter's vote(s) for one of the alternatives being considered.

5. Add the vote(s) inputted to the tally for that alternative.

6. Check if there are any more voters? (This decision can be made by pre-setting a deadline for voting or by pre-setting a total number of voters.) If yes, go to 3 above. If no, continue.

7. Tally N, the total number of votes.

8. From the set integers, 1-N, assign to each alternative a set of winning integers, the set being equal to the number of votes tallied for each alternative.

10. Get a random integer from 1-N.

1 1. Find the alternative that has an integer assigned to it that matches that random integer.

12. Output that alternative as the winner.

Authentication Method and System Using Double Purpose, Intuitive Passcodes

We have three parties: Bob, Merchant and Bank. The idea of the Neutral Party Passcodes is that the neutral party (the Bank) issues a Passcode to a first party (Bob). The Bank keeps that code as well, secure and locked up. There is nothing new about Passcodes. They verify identity.

The method has two key ideas in addition to the lock-box Passcode principle. The first idea is that the Passcode not only verifies identity but can verify that other messages are genuine. The second idea is that this coding is easily understood by the public and easily used without encryption apparatus. It is a simple coding method that anyone can understand and use and that comes naturally as well.

We will now describe the basic codings and discuss their operation within a lock-box, Passcode system.

1. Words can stand for letters such that the first letter in a word stands for that letter (e.g. "Blue" = "B").

2. The number of letters in a word can stand for the corresponding counting number (e.g. "Blue" = 4).

3. A word can stand for a number according to a personal cipher (e.g., a person's cipher might have "knot" = 1 , "jello" = 2, "Smith" = 3...).

The messages one can code for are endlessly diverse. Let us stick right now with a very important type, the type of message useful for the transferring of money. Our goal is to send messages over non-secure channels such that the messages allow the transfer of funds.

The Bank issues Passcodes to Bob. These consist of words (Note: Bob could pick the words, as in picking a PIN number, but we will ignore that not very relevant variation). Bob could get a full alphabet worth of words (he could just get specific words as needed but here we assume he gets a full alphabet).

Apple = A, Bike = B, Cull = C, Dog = D....Girl = G, Hill = H.... Nancy = N.... Rip = R....Vic = V, Wiper = W....

These words would not only be registered to Bob but could also be associated with payment packet amounts. We assume for this illustration that this alphabet is associated with $ 10 packets. (Interestingly, an easy way to denote the amount of the packet is to use the number of letters in the words. Thus four letter words might signify 1000 pennies.) Now Bob would have these words and Bob would have to present a payment message to the Merchant. Let us assume that Bob is using credit packets. And so Bob presents a message saying, I have $10 credit and here is my Passcode "Cull." The Merchant calls the Bank and says, the Passcode is "Cull" and the user name is Bob. The Bank checks and sees that Bob indeed does have "Cull" registered and sees that Bob does have $10 credit associated with the set of words that "Cull" is in. And so the Merchant accepts Bob's "money."

Of course, the Bank can debit Bob's account and cancel the word "Cull" from the list if the words are being used as money themselves, rather than as indications of credit. The words then become one time pads such that each word can have a dollar amount associated with it, and once used, the word gets canceled from Bob's list.

Now a potential problem with sending such a message over an unsecured channel is that someone might steal the packet information and present it elsewhere as payment. This thief would be pretending to be Bob and another Merchant would not know any better.

And here comes the great trick. Using the words-alphabet cipher to encode an extra message. Instead of using the word "Cull," Bob can use letters that identify who the receiver of payment is supposed to be. The principle is the same as making out a check to a particular person instead of a blank check. For example, say the merchant is Banana Republic, Bob could use as a Passcode the word string "Bike, Apple, Nancy." Now the Merchant could indeed verify these with the Bank. But more important, if a thief was to steal the credit packet, he would have to spend it at a vendor with the beginning letters of BAN. (Of course, which letters are chosen is not the point. The method can be standardized.)

Claims

Claims

I claim:

I . a special purpose counting machine (CM) including means for entering, storing, manipulating and outputting data, that executes the steps of:

pre-storing data that identifies the ad whose audience is being counted, and storing the matching message (MM) that is associated with the ad, and storing a multiplier associated with the MM, setting a tally to 0, setting an audience count to 0, and when receiving a response from an audience that sees an ad,

—registering the ad ID data,

—registering the MM in the ad,

-adding to the tally of responses for that ad,

—pulling the MM multiplier,

-multiplying the tally by the multiplier to yield the audience count,

—outputting the count.

2. the machine of claim 1 , which additionally executes the steps of:

-storing a table of correcting factors (CF's) that correspond to pairings of demographic categories and reward amounts and, when receiving a response from an audience that sees an ad also

—registering ID data for the responder,

—registering the MM for the ad,

—registering the reward amount for the MM,

—registering the demographic category for the responder,

—pulling the CF that corresponds to the pairing of the reward amount and demographic category registered,

—applying the CF to the MM multiplier for that response,

—multiplying the corrected MM multiplier by 1,

—adding the MM multiplier to the audience count.

3. a computer including processing means, memory means, input/output means, and display means that executes the steps of: a) inputting information identifying the case being considered, b) inputting the alternatives being considered, c) inputting N, the total number of votes to be cast, d) inputting the votes cast for each alternative, e) registering the number of votes cast (the probability weighting) for each alternative, f) from the set integers, 1-N, assigning to each alternative a set of winning integers, the set being equal to the number of votes registered to each alternative, g) getting a random integer from 1-N, h) finding the alternative that has an integer assigned to it that matches that random integer, i) outputting that alternative as the winner.

4. a computer including processing means, memory means, input/output means, and display means that execute the steps of: a) inputting information identifying the issue being considered, b) inputting the alternatives being considered, c) inputting a voter's ED information, d) inputting the voter's vote(s) for one of the alternatives being considered, e) adding the vote(s) inputted to the tally for that alternative, f) checking if there are any more voters and, if yes, going to 3 above, and if not, tallying N, the total number of votes, g) from the set integers, 1-N, assigning to each alternative a set of winning integers, the set being equal to the number of votes tallied for each alternative, h) getting a random integer from 1-N, i) finding the alternative that has an integer assigned to it that matches that random integer, j) outputting that alternative as the winner.