FAQ

home

What is Logicracy?

Logicracy is a free, on-line process that gives the people the power to solve some of the most profound social and political problems facing many nations today.

In a nutshell, Logicracy makes it possible for the people to easily define and implement a decision-making process which the majority can agree will produce much better decisions than are currently made by any government, and use that process to produce decisions which can then be compared with the decisions made by government. When a decision made by the process differs from the decision made by government, the government's incompetence is proven. It isn't necessary for people to agree on the "best" decision making process. They need only agree the process is better.

For example, imagine that a surprise question is given to voters as they leave the voting booth, like "Who was buried in Grant's Tomb?". By eliminating or reducing the weight of the vote of those that get that question wrong (the bottom three percent), the outcome of many close elections would be changed. Most would agree this decision making process is better than our current process--thus its outcome can be used to test the outcomes of our current process. Logicracy provides analysis tools and data gathering techniques that make a huge variety of such decision-making processes possible, including the one described here. Logicracy also makes it possible to eliminate all but a negligible amount of fraud from the online data gathering process.

Logicracy is non-profit, there are no advertisements, and membership is free.

Right now Logicracy is brand new and there are very few testing questions and referendums. These must be submitted and judged for quality by the people in order to solve the problem of bias. Many more of these questions are needed to get Logicracy off the ground and producing useful results. If everyone waits for Logicracy to gain more questions, more users, and become better known, then it never will. Please consider joining, making suggestions, and submitting new skill-questions and referendums to your favorite subjects.

Is Logicracy better than democracy?

This really depends on your definition of "democracy".

If you think democracy is the process of being allowed to vote only for one of two candidates nominated by parties whose main criteria in nominating a candidate is his or her ability to muster campaign money, his or her looks and speaking voice, how tall the candidate is, and finally whether the candidate agrees with the party platform, then yes, Logicracy is better than democracy.

How does Logicracy work?

• Opinions on many different user-submitted referendums
• Ability to predict the future (by comparing a user's previous prediction with present fact)
• Verified geographic location (from non-proxy network address) when known to be reliable. This locates the user to a region, sometimes a city.
• Absolute mental effort put forth
• Ability to estimate one's abilities
• Time frame in which a referendum was answered
• Various indicators of user character
• Consistency in answers to sets of opinion questions
• Skill and knowledge in any of 160 different subject categories. This can include, for examples, esoteric knowledge of a geographic region's events and attributes, or skill and knowledge in subjects related to the referendum, like, for a referendum on global warming, a combination of geophysics, politics, and the politics of scientific publication and funding.

Here is the process:

1. Several internal mechanisms in the site are used to minimize forms of fraud like account selling, vote selling, foreign voting, bots, repeat voting, and other forms of vandalism. Analysis, as described below, eliminates the lion's share of what remains.
2. Users submit and judge skill-testing questions and referendums in many different subject categories. Only questions judged by the people as high quality will have a notable effect on statistics and user expertise in a subject.
3. Each user chooses categories from which skill-testing questions will be taken, and categories from which referendums will be taken. Questions are randomly chosen from the categories and given to the user to answer and judge for quality and subjectivity. Referendums are given less frequently than skill-testing questions. This tests the user's skills and helps keep repeat voting under control (the effect of repeat voting is further reduced in analysis as described below) . Each time a user answers a referendum, everything about the answer is recorded--like user's recent skill rating in each category and his answer and judgment of the question.
4. Any user can then analyze a referendum by creating a scatter plot of votes as shown below. Each point is a vote on the referendum. The horizontal axis is typically defined as the voting user's opinion on the referendum and the vertical axis the user's "competence". Competence can be defined as any mathematical combination and degree of user attributes like those listed above.
5. One can tell much from the scatter plot, including (usually) the result of the referendum and even evidence of any notable fraud. Notice that repeat votes from the same user will simply appear as more points at roughly the same location in the chart. Also note that repeat votes with high competence indicate more effort was exerted by that user--which some analyzers might choose to interpret as at least a partial indicator of character. Remember that repeat voting on a given referendum, while the user's expertise is high, requires that user to exert mental effort to keep his expertise rating high.
6. Further analysis might be done. (Currently, Logicracy provides the tools to create the scatter plot. The following processes require the data to be downloaded and analyzed locally).
7. Outliers can be removed from the distribution using whatever algorithm the analyzing user chooses. For example, this may be as simple as a lower error bar to eliminate low-competence repeat voters. In any case, remember that outliers and extreme repeat voters will be fairly obvious in the scatter plot.
8. A curve can be fitted to the distribution. This might be an upper bound curve if desired, to further eliminate low-competence repeat voters. If the curve-fitting algorithm uses a density limit, then the effect of repeat voting is again notably reduced and all that is left is a slight weighting of votes by mental effort exerted and competence as the analyzing user defined it. The maximum of the curve could be interpreted as the referendum result.

Here is an example scatter plot. Notice the massive low-competence repeat votes in the lower left corner. This was obviously an attempt by a user to pull down the left side of the curve. It was easily removed as a low-competence outlier below the error bar. And even if it hadn't been removed, viewers of the chart would know what it was, anyway. Note that repeat high-competence votes require notable mental effort and, to some extent, character. The upper bound curve fitted to this distribution indicates this referendum should not be implemented.




Logicracy also...

• Provides regional and local categories where you can submit and answer questions to test local knowledge and vote on local issues. Post a request in the 'Site design' forum if you would like your town, city, or region/state added (please note, however, that local categories with very low traffic for extended periods will be deleted).
• Provides private groups to select organizations. These groups are accessible only by use of a password. Such private groups might be useful, for example, for a political science project or to make decisions within a private organization. To request a private group, first join Logicracy, then go to the forums and send a private message to user 'admin1'.

How does one pronounce 'Logicracy'?

lä ji ' kru s e '

Click here to listen.

Isn't Logicracy based on the questionable notion that people with high I.Q. will be better leaders?

Logicracy does not define competence, or even the analysis technique you use to make decisions. That is up to the user doing the analysis. One of the main purposes of Logicracy is to discover many correlations rather than presume any one of them.

Doesn't Logicracy really just implement competence-weighted on-line democracy?

Analysis can be that (as long as mental effort is included in the definition of "competence") and it can be much more. Consider the following differences between an on-line democracy and Logicracy:

• There is no such thing as an on-line democracy. An on-line democracy cannot be implemented without severe vulnerability to various forms of voter fraud. Logicracy can be practically implemented on-line because it is immune to any substantial effect from such fraud--or at least allows such fraud to be measured and removed in analysis.
• Democracy still requires some elite group to compose referendums. In Logicracy, the text of all referendums are composed by the people and, using typical analysis, judged for quality by each citizen weighted by his competence and character. Among other things, this discourages spin (clever wording of referendums for ulterior motives) and riders (unrelated prospective laws attached to a referendum--like pork, which is allocation of public funds for operations everyone knows would be against the will of the majority).
• Logicracy is to quality as democracy is to quantity. Besides the myriad correlations that can be revealed by analysis of the raw data, Logicracy can make decisions by fitting a curve to data points in a space of competence vs. opinion rather than numbers vs. opinion. The slope of the high-competence end of the curve indicates the decision even if there are fewer votes there (as long as they passed your outlier removal filter). This is not the same as even a competence-weighted democracy.
• Logicracy presumes no correlation of any citizen attribute with leadership ability, although typical analysis should consider at least some degree of absolute mental effort in order to eliminate fraud--see details about that elsewhere in this FAQ. Democracy presumes a correlation between majority and leadership ability, and nothing else.

How is Logicracy kept free of bias?

All of the material used by Logicracy to measure user attributes is submitted and judged for quality by all the people. When you answer a skill-testing question, the effect it has on your skill rating is directly proportional to its rated quality. Only the skill-testing and opinion-testing questions that have been judged well by the people have notable effect. The only "bias" in the system comes from all the people.

How can Logicracy possibly measure character?

User character can be measured in four ways:

1. Analysis can account for a user's willingness to contribute mental effort in the definition of "competence". In such analysis, the more absolute mental effort put forth--specifically, the more skill-testing questions the user answers well--the more power the user can have in Logicracy. The degree this affects a user's competence rating depends on the type of analysis that is chosen.
2. Users estimate their own abilities by specifying how sure they are of answers to skill-testing questions. A user that overestimates his abilities will lose more on a question when he answers it poorly.
3. Users are charged very slightly more skill points for answering an opinion question differently than are most other people. A person of lower character is more likely to answer the same as others so that he doesn't lose as many skill points.
4. Users can choose whether to make their skill ratings visible. A user can forego the reward and fame of having a high published expertise for the sake of improving his credibility in analysis of raw data--assuming the analyzer chooses to account for it.

Are you saying that the more time a person spends in Logicracy, the more power he has?

This is the case only if democratic or weighted-democratic analysis is used. You can do that analysis if you want, but the best analysis is that analysis that is unaffected by number of votes--a competence vs. opinion scatter plot.

In a competence vs. opinion scatter plot, adding more points to the low-competence end of the scatter plot only serves to make the slope of a line fitted to the scatter plot more obvious. Such a scatter plot requires only enough high-competence (and low-competence) votes to overcome any outlier removal process. That's the whole purpose of fitting a curve to such a scatter plot.

The main purpose for requiring users to exert mental effort for the opportunity to vote is to eliminate the danger of a bot answering opinion questions ad infinitum. It is a pretty good solution. Not only does it virtually eliminate the danger of bots (see the question on bots ), it further weights the quality of votes according to the intelligence and perseverance of the voters. Yet this weighting is only slight when analysis of a competence vs. opinion scatter plot is done.

Is Logicracy vulnerable to the fraud of voters placing multiple votes or assuming multiple identities?

Splitting one's mental effort among multiple accounts does not increase the number or competence of votes added.

The combination of the following make the effect of repeat voting negligible:

• Logicracy detects rapid responses to questions from a given IP address, and substantially slows them down.
• It is easy and usually useful to remove outliers below a certain competence level.
• IP address and IP subnet can be considered in the analysis of a given referendum. For example, only a small number of votes from each IP address could be included in the data set.
• A data set can be taken from a short period of time in which no voter could vote on a given referendum more than a few times.
• The handful of extra votes a user can make within the above limitations simply places more points in the same position in a competence vs. opinion scatter plot. Outlier removal and density-limited curve fitting can remove even more repeat votes.
• There is generally some correlation with voter character and amount of mental effort put forth. It is reasonable to give a higher-character voter at least a little more power than one of lower character.

Of course, analyzers and users criticizing an analysis should consider many things when judging the reliability of an analysis, like the number of unique IP addresses in the data set, the number of votes, and the number of opinion questions in the subject--which is one reason the user's current opinion subject is recorded in the raw data.

Is Logicracy vulnerable to automated client "bots" posing as humans?

Many interactive Internet sites suffer from abuse by "bots", which are client-side computer programs masquerading as humans. Sites often employ tests for (hopefully) discriminating bots from humans. Such a test is called a captcha (Completely Automated Public Turing test to tell Computers and Humans Apart). The best captcha must be designed to maximize the ratio of a human's performance to that of a machine, and the tests must be given frequently and require prompt answers so that a reliable ratio is measured and so that bots can't conveniently relay hard tests to groups of real humans that were somehow compensated for their help. Since Logicracy just happens to do exactly that--require prompt, frequent, answers to questions that require thought--Logicracy is a natural captcha of the most reliable kind.

Nevertheless, for added security, users must answer a simple (one-mouse click) image captcha for each question asked. This is a distorted image of text that must be resolved by the user. To answer the image captcha, the user must also have read the question (because the user is also asked whether the word in the image is from the question text or not). For more on this, see the question about the image captchas .

The best effort at a bot design would require...

1. An up to date and comprehensive database of the right answers to the vast majority of skill-testing questions.
2. A human on call the entire time that still must help with answering the image captchas well.

If the bot's database of right answers is not comprehensive (possessing at least 90% of all questions and their right answers), then the bot must either relay hard questions to a human, or it must appear as a low-competence human and simply contribute to the low-competence end of the competence vs. opinion scatter plot.

A comprehensive database of right answers cannot be created because once there are enough new questions being submitted each day, older, lower quality questions can be retired and the question turnover rate will outrun the ability of even a group of humans to a keep an up-to-date database of questions and their right answers for use by a bot. The human effort involved in keeping an up-to-date database with right answers (often requiring research) and in answering the captcha (by either reading every question or typing in the captcha text), will be less effective than simply letting those humans answer questions normally. This brings up the issue of vote selling . Also see the question about users compiling a local list of questions .

What about users buying and selling votes?

The combination of several deterrents to vote buying should make vote buying effectively nonexistent:

1. In a raw democracy, a vote buyer need not be selective in finding voters to do his bidding because every vote in a democracy counts the same. When decisions are made according to a scatter plot of competence vs. opinion, however, the vote buyer must be very careful that only users with the very highest competence are hired (he must have some method of measuring this), or he will not affect that upper end of the competence curve. So the vote buyer must reliably find a substantial number of notably high-competence users that also have notably low character and that also happen to need money desperately.
2. Certainly there are regions on the Earth where such high-competence, low character, poor users exist--but they must also pass any IP address geolocation filtering being used in the analysis.
3. A vote buyer must advertise to find those poor, high-competence, low character users in a way that does not reveal his intentions too publicly.
4. A vote buyer must compensate his hirelings enough to make it worth their while and not worth exposing the scam.
5. There must be enough hirelings to overcome the outlier removal process in analysis. Note that a sold vote is more likely to be an outlier.
6. Sold votes may, in some cases, be visible in a competence vs. opinion scatter plot as a concentration of outliers.
7. The effect of the vote sellers must be so formidable that it fully changes the result of a referendum.
8. Logicracy allows users to hide their performance  from vote buyers and their bots so that a vote buyer can't tell whether his hireling is doing his bidding without candidly watching him individually for an extended period of time (and knowing the right answers to the skill questions being answered). Analysis can give hiding users more weight. Analysis can also use the difference in the statistics of hiding users and non-hiding users in order to estimate the degree of vote selling and compensate for it.
9. The vote buyer must develop software to run on the hireling's computer that let's the vote buyer monitor the hirelings progress. This software must be immune to hacking by the vote sellers and be kept updated to handle changes in the site interface.
10. If it is suspected vote buying is a problem, Logicracy could simply regularly make changes to its interface that would counteract the vote buyer's bot. In an extreme case, users could be given the option to have the question text presented in a way that is easy to resolve by humans but difficult or impossible to resolve by a bot. For example, it could be presented the way the captcha text is presented, but less distorted. Analysis could give the answers by such users more weight, or even use the difference in the statistics (between that of users using this method and not using the method) to estimate and compensate for vote buying--similar to the way hiding can be used in analysis.

What about users buying and selling accounts?

Each time you answer an opinion question, only your recent performance in answering skill-testing questions is recorded for analysis. So even if people do buy and sell accounts (which is, by the way, a violation of the terms), any performance of the previous owner of an account will last only over the next few opinion questions, and it is unlikely any of those opinion questions will happen to be the ones of great interest to the account buyer (since they are randomly chosen). Also, even the recent skill rating that can be stored must still be visible to be sold, and there is an option to hide your skill rating (even from yourself) in order to increase your effect in Logicracy and deter vote buying in the process.

Might there be clever strategies for a user to usually answer skill-testing questions right even though he doesn't know the answer?

Skill-testing questions that can be answered easily, whether because they are inherently easy or are subject to some clever answering strategy, will be rated as lower quality and/or easy and will thus have less effect on the answering user's skill points, and likewise on statistics. A user following some clever answering strategy in hopes of often getting right answers without mental effort will only succeed in answering these "easy" questions moderately well and still fail to correctly answer the more difficult questions which have more effect on his competency. That user will either contribute to the low end of the competency curve in a scatter plot (simply making its slope more obvious), or be eliminated as an outlier in analysis.

What about users compiling a local list of all questions?

The site provides several ways for a user to search for and view skill-testing questions or opinion questions. Having quick access to lots of skill questions doesn't really help a user or affect statistics. It is the correct answers to those questions that make the difference. See the question about bots for further discussion about this.

Nevertheless, each of the several methods the site does provide for accessing questions contains a limitation to hinder abuse. For example, in some cases AP (analysis points) are charged. Also, a user creating a "throw-away" account for viewing questions en masse will find that he must still at least answer the questions and captchas well or the rate that he can view questions slows down dramatically (this is done per IP address).

What about users that answer questions en masse with wrong answers in order to skew the "right" answers of questions?

Since a question's "right" answers and quality and subjectivity ratings are resolved from the way high-skill users are answering the question, some users may attempt to force the "right" answers, quality, and subjectivity of questions to a known state. This might be attempted so that when the question is asked again, skill points will be won. However, it may also be done out of maliciousness.

This will not work for the following reasons:

1. The skewing must be performed by a human and not a bot, because the captcha image must still be answered.
2. To have a notable effect on a question's quality rating and its right answers, you must have a notable skill rating when you answer the question. To gain skill, you must answer many previous questions correctly.
3. To appear in the high-competence end of a scatter plot, a user must answer questions notably better than the average user. This means the skewed answers must be more than just skewed. They must be practically set outright. To do so would take far more time and connections than are physically possible. Each question would have to be answered many times while having a high skill rating. The question turnover rate would never let this happen. A bot attempting to do this rapidly would be immediately detected and slowed to a crawl.

Will a new subject with few questions be vulnerable to vandalism?

One can imagine a malicious user adding many bogus questions to a new subject. Since this is a violation of the terms, the administration reserves the right to remove such questions. The total number of questions removed from a subject will be published. In that way users can decide for themselves whether there is a possibility that the deletion of questions has compromised the statistics.

What about hackers, or even site administrators, tampering with the data?

Hacking of the server is certainly possible. The security mechanisms used by the Logicracy server are reasonable and fairly standard, but no guarantee can be made that the server won't someday be hacked. The database is backed up fairly regularly. So at least if a detected breach occurs, the database can be restored to a recent state. If Logicracy grows to great importance, the same security that on-line banks use can be employed in the site. Hopefully those that trust on-line banks with thousands of their dollars will trust their votes to the same level of security.

It is important to understand, however, that security of individual accounts is not that important. Logicracy is designed so that theft of a password will not adversely affect the integrity of the statistics and referendum results produced by Logicracy (see this question ). For this reason, not all of the security of an on-line bank will be needed even if Logicracy does grow to importance. For example, on-line banks use HTTPS network connections to reduce the probability of one type of password theft (packet sniffing), and they often use graphic password entry to help reduce another type (keylogging). Logicracy needs neither of these except as a convenience to users (because nobody likes having their password stolen). Since HTTPS can degrade performance substantially, HTTPS is not used.

Attempts were made while designing Logicracy to somehow cause any manipulation of data by the site administrators to be immediately detectable. Unfortunately one can always imagine a form of manipulation that would be impossible to detect by users. For example, a bot that has access to the Logicracy database could easily emulate a user of high competence. There are ways to fight the danger of internal fraud if Logicracy were ever used for serious decision making. Independent groups could have the ability to monitor, but not alter, the internal operations and accesses of the site, or multiple sites could be administered by independent administrators and the statistics gathered from those sites compared. Note, however, that these are also just as much a danger in current implementations of democracy.

What about users doing research before answering difficult questions?

You are welcome to do research before answering a question.

Consider:

• Research improves the reliability of the question's calculated "right" answers.
• It indicates that the user is intent on having an effect in Logicracy--which is somewhat a measure of character.
• Doing research does slow down the rate that a user can answer questions. Users that know answers without having to do research have the advantage of answering questions more quickly.

Why doesn't a local, regional, or national use of Logicracy need voter registration and verification in order to eliminate foreign influence?

Your outlier removal process and your definition of voter competence can filter the lion's share of foreign influence from any data. For example, you can base them on a combination of the user's non-proxy IP subnet, knowledge of esoteric events and facts pertaining to the user's claimed region, and language skills. In some cases those filters could actually be better than voter registration, for voter registration, to some extent, can give more power to local traitors and incompetents at the expense of competent foreign patriots.

Even so, at some point voter registration might be added to Logicracy. It would, of course, be difficult to make secure and it would likely require notable administrative expense. In fact, there would need to be even more independent oversight of the administration of it--for a corrupt administration of it would make things far worse than they would be with no voter registration at all.

What is intentional skewing and how is it overcome?

Intentional skewing occurs when a user masquerades as a member of a group of which he is not a part, and intentionally answers questions in a way that would slander that group.

For example, a member of one political party might log in and answer political opinion questions as would a member of the opposing party, and then intentionally answer intelligence questions poorly in the hope of causing some statistics to indicate members of the opposing party are unintelligent.

It turns out that this ploy is always extremely counterproductive for that user and his agenda for the following reasons:

1. The user has sacrificed the time he could have spent promoting his own agenda as a weightier user and having a (probably) greater overall effect.
2. In a democratic analysis of the raw data, the user is simply promoting the agenda with which he disagrees.
3. In a competence-weighted democratic analysis, the user is still simply promoting the agenda with which he disagrees.
4. Competence vs. opinion scatter plots that do not show a clear curve naturally reduce to competence-weighted democracy, anyway.
5. In a competence vs. opinion scatter plot, even when a curve can be clearly fitted, the masquerading user must precisely adjust his competence so that he does not become an outlier. In fact, it is the slope of the high-competence end of the curve that is most revealing. A masquerading user must precisely adjust his competence to lie just below the high-competence end of the curve. If he doesn't, then he either promotes the agenda he disagrees with, or his efforts are ignored as outliers and he wastes time that could have been used to further his own agenda.

The bottom line is that intentional skewing actually contributes to Logicracy's usefulness--it causes users of low character and intelligence to inadvertently promote the agenda they disagree with at the expense of their own agenda.

What about people who don't have Internet access?

Let's say two people are presented to you. All you know about them is that one uses the Internet and the other does not. You must pick one of them to choose whether the nation should go to war or not. Which one would you pick?

How are malicious question submissions handled?

A very low quality question will be judged so by the first two or three users that see it. After a question has been judged by several users it is deemed "mature", at which time it will be automatically disabled or deleted if the quality is substantially low in most of its subjects.

See the FAQ question about question submission for more information on what is acceptable in a question.

What about obsolete referendums?

The quality rating of a question reflects only the last few hundred user quality judgments. A referendum or even a skill-testing question that, for one reason or other, becomes obsolete, will fairly quickly lose its quality.

What are all these points, levels and ratings?

The current values of your various points, levels, and ratings are visible on your home page. Some of these are kept separately for each subject (so you will see on your home page only the parameter's value for the currently selected skill-testing subject), others are not associated with a separate subject (the parameter's value shown on your home page is unrelated to the currently selected subject).

Only your Skill Rating is used in analysis. So it is, of course, the most important parameter. See this question for more details. Here are descriptions of the different parameters.

Skill Rating:
You have a separate skill rating for each subject. Your skill rating is an indication of your recent performance in a subject. If you have answered all of the skill-testing questions exactly correctly in the subject, then your skill rating is 1. If you have answered all of them exactly incorrectly, then your skill rating is -1. Each time you answer an opinion question, your current skill rating is recorded in that answer so that it can be used in analysis.

Expertise level:
Your expertise level, or 'level' for short, is just for show. It appears in your forum posts and on your profile page. you can brag about it if you like, but that's all you can do with it. Your expertise level is an integer proportional to the square root of the total skill rating you have "earned" from each skill-testing question you have answered.

Question submission points (QP):
QP is kept separately for each subject. You earn QP by answering skill testing questions well in the subject. You spend QP when you submit a question .

Analysis points (AP):
Your AP is kept in a single bucket (it is not kept separately per subject). You earn AP by answering skill testing questions well in the subject. You spend AP when you search for questions and when you perform an analysis .

Forum points (FP):
FP is kept separately for each subject. You earn FP by answering skill testing questions well in the subject. You spend FP when you post a message to a forum associated with the subject.

User judgment points (JP):
Your JP is kept in a single bucket (it is not kept separately per subject). You earn JP by answering skill testing questions well in the subject. You spend JP when you use the "Judge user" function. [Note: The full purpose of JP is not yet defined].

How does Logicracy calculate my skill and how does my skill affect opinion questions?

First, remember that skill in a subject is only one of several user attributes that can be used in analysis. If you don't believe Logicracy can reliably measure skill, then don't use skill in your analysis.

Skill rating is not to be confused with expertise level , which is not used in analysis.

You earn a skill rating in a subject by answering questions in that subject. Before you start answering questions, you specify the subject from which skill-testing questions will be randomly taken and the subject from which opinion questions will be randomly taken.

You are given about one opinion question for every ten skill questions. A new user, however, will not receive an opinion question until after the first fifty skill questions.

You may answer any number of questions any time you like.

Your skill rating is kept separately for each subject. It is an indicator of your recent performance (over roughly the last several hundred questions) in a subject. Your skill rating is 1 if you answer every question exactly right. Your skill rating is -1 if you answer every question exactly wrong. Your skill rating for each subject is recorded in the raw data each time you answer an opinion question. Skill rating is typically used in analysis.

Your skill rating in all subjects is slightly decreased each time you answer a question. So you must periodically answer questions in a subject to keep a high skill rating in that subject.

Questions are multiple choice.

You can also submit new questions. When you submit a new question, you do not specify the question's right answers, quality, difficulty, or subjectivity (how much it tests opinion as opposed to skill). The system resolves all of these according to how people answer and judge the question. (The higher a user's skill, the more effect that user has on the resolved 'right answers', quality, and subjectivity of the question).

When you answer a question, you specify separately how much you agree (on a scale of 1 to 5) with each optional answer, how sure you are of your answer, your judgment of the quality of the question in the current subject, and your judgment of the question's subjectivity.

Obviously, skill-testing questions and opinion questions work differently:

Pure skill-testing questions work this way:

• Your gain (or loss) of skill rating depends on how closely your answers and quality and subjectivity judgments match those of the question.
• The degree of your gain or loss depends on the product of how sure you say you are and on the quality of the question.
• The question's right answers, quality, and subjectivity are calculated from an average of each user's answers weighted by the product of his current skill rating and his sureness specification. Question quality is kept per subject, so it depends on the per-subject skill rating. Right answers and subjectivity are altered proportionally to the product of the quality of the question in the current subject, the user's skill in the current subject, and the user's sureness.

Pure opinion-testing (subjective) questions work this way:

• You neither gain nor lose skill rating when answering an opinion question.
• All details of your response to the question, including how long you took to answer the question and your skill rating in each subject, are recorded as raw data for later analysis.
• The question's quality and subjectivity are an average of each user's answers.

The higher a question's subjectivity, the more it is treated like an opinion-testing question and the less it is treated like a skill-testing question.

A newly submitted question is marked as "immature" and given a very low quality and a subjectivity that is .5 (half skill-testing half opinion-testing). Even though the quality rating is low, an immature question has a high probability of being asked of you. As the question becomes more mature, it will be treated more and more like a normal question.

You also earn other types of points by answering skill-testing questions well. These points are earned according to how well you answer skill-testing questions:

• You earn forum posting points (FP) when you answer questions well. These are kept per subject and can be spent to post to that subject's forum.
• Question submission points (QP) are kept per subject. You spend QP when you submit a question to a subject. When you submit the same question to multiple subjects, you are charged QP for each of those subjects. Question submission is quite cheap. In fact, it is free for subjects that don't currently have very many questions. The main purpose of regulating them with QP is to hinder malicious flooding of the system with low quality questions.
• Analysis points (AP) are spent when you download and analyze data and when you search for questions.

Tell me about hiding my skill rating. Why would I want to do that?

(Note: hiding is not yet implemented in the site)

Since many users will want to know and publish their skill, the hiding of one's skill rating is optional. A user may turn on or turn off the visibility of skill rating whenever he wants.

The hiding of your skill rating (even from yourself) is a useful tool in deterring vote selling, account selling, and it can increase your effect in Logicracy because some analyzers will account for it as an indicator of higher character.

The number of consecutive hidden answers by a given user is recorded in the raw data when that user answers an opinion question. Analysis, then, can weight answers according to how long a user has been hiding. Analysis can also use the difference in the two types of data (hidden and exposed) to estimate the degree of vote buying that is taking place, and compensate for it. By not hiding skill points for a time, a user can create a visible skill rating for himself that will appear in his forum posts and profile page, for example. (The skill rating only reflects the user's activity while not hiding his points). By hiding skill points, a user can deter vote buying and even increase his weight when an analyzer of raw data decides to assign more competence to those hiding their skill points on the assumption that such users have higher character and are less likely selling their votes.

When a user stops hiding, his visible skill rating is automatically set to the skill rating he had when he started hiding. In this way one cannot tell from the user's current skill rating what the user's skill rating was when opinion questions were answered while hiding. Of course, performance would be apparent if a user hides for only a short time, but reasonable analysis would not consider users hiding for a short time notably different from users that were not hiding. However, if non-skill points ( AP, QP, and FP ) were still accrued normally while a user is hiding, then they could be used to indicate a user's performance. For this reason, while a user is hiding, the rate AP, QP, and FP accrue is randomized by 30%. This is necessary to prevent a user's performance from being resolved at the moment he or she stops hiding, but still allow those points to be accrued. When a user begins hiding, a scaling factor is chosen which varies between 70% and 130%.  AP, QP, and FP are accrued according to this factor while the user is hiding. When a user stops hiding, one cannot tell from that user's AP, QP, or FP what the user's performance was, except within 30%. The scaling factor, of course, is never revealed. (Since most analysis is concerned with the slope of the high end of a curve fitted to a competence vs. opinion scatter plot, that 30% variation totally hides a user's performance in that important range of competence). In this way hiding users are not generally penalized in AP, QP, or FP while they are hiding.

Reasonable analysis should probably not consider hiding users much different from non-hiding users until they have been hiding for at least several hundred consecutive questions, This is because by then a user's performance will be fairly well hidden when that user stops hiding.

Whether a user specifies his skill points should be hidden or not, several other techniques are used to hide certain information in raw data no matter what:

• User IDs in raw data are encrypted, and encrypted differently in each data set.
• Data sets containing data for less than 10 users will not be allowed.
• The data set ending date is secretly randomized by up to a week.
• Only opinion questions can be analyzed.
• When a user gathers and analyzes data, that user must spend analysis points (AP). AP can be earned only by exerting mental effort. The points of a given hiding user could be resolved in spite of the efforts to hide them (through clever analysis) only by spending a tremendous amount of AP. Since user's cannot continually gather such large amounts of data (because they would run out of AP), the skill points of a hiding user cannot be resolved in this way.

How do I report abuse?
How do I judge another user favorably or unfavorably?

(Note: The process and purpose of judging users is not yet fully defined)

You can use the 'Judge User' page to judge another user favorably or unfavorably on a whim, or to report a violation of the terms (even if you don't know the name of the user).

Judging a user does not affect the judged user's skill rating in any way--your judgment of a user does not affect statistics in Logicracy. (Unless, of course, you report a provable violation of site rules which results in the banning of the judged user).

You can get to the 'Judge User' page directly from your home page or you can go to a user's forum profile page (via one of that user's forum posts or the forum member list) and then click the 'Judge this user' link at the bottom of the user's forum profile page.

One of the types of points you earn when you answer questions is judgment points (JP). To judge a user or report a violation of the terms, you must specify on the Judge User page at least some JP to spend. You'll not be able to judge or report a user until you have at least some JP. New users start with a negative balance of JP--so it will take you a while answering questions to get a positive balance of JP.

You can use the Judge User page to judge a user favorably or unfavorably on a whim. That is, you don't have to have a good reason to do so. For example, you might decide to judge a user unfavorably because you disagree with a forum post by that user, you don't like the user's user name, or you find the tilt of his hat a bit too ostentatious. In that case, DO NOT place anything in the comment box at the bottom of the Judge User page. The comment box is strictly for reporting violations of the terms.

To report a violation of the site terms, specify at least some JP, specify an unfavorable judgment, and place a comment in the comment box that describes proof of the violation of the site terms.

The amount of JP you specify is used to alter the JP of the judged user. For favorable judgments, the judged user is awarded one-half the JP you spend. For unfavorable judgments, the judged user is fined one-fifth the JP you spend.

If you encounter a user behaving badly, please note the following rules before taking any action:

Never submit any information to the site that could reveal a suspected offender, except as an abuse report as described herein. You can civilly discuss an offence in, for example, the forums, as long as the discussion complies with the terms and forum rules and no hint is given about the identity of the offender.

Please use the comment box on the Judge User page only for violations of the terms that can either be proven or are occurring frequently, and may not yet be known by at least one moderator or Logicracy staff member that should know it. If it isn't a direct violation of the terms, it can't be proven, or you think a moderator or staff member that should know it most probably already knows about it (except for a moderator that violated the terms), then do not submit a report about it with text in the comment box.

If there is a behavior that you think is bad, but is not addressed in the terms, do not include text in the comment box. In that case you are still welcome to judge the user unfavorably by using the Judge User page (without specifying comment text) and you are welcome to civilly discuss any disagreement you have with the terms or site design in general, as long as the discussion complies with the terms and forum rules and no hint is given to the identity of an offender.

In a terms violation report (the text in the comment box), be sure to include everything you know of (or information leading to) the following: user name of the offender, time of offense, what part of the terms they violated (like a short quote from the terms), and details of the offense. For example, if you see a question that violates the terms, be sure to include the question number. Also, please be concise.

Do not report (with text in the comment box) more than once for a single instance of a given abuse. Also, moderators are instructed not to reply to the user submitting an abuse report unless they need to ask for more information.

This reporting process may be updated periodically or even replaced. So be sure to check this section before each time you send an abuse report.

What do I need to know before using the forums?

Among other things, the terms list all the things you are not allowed to submit anywhere in the site. These include, but are not limited to, slander, lies (except as an optional answer to a question), abusive material, illegal material, even the most mild profanity, material designed to get around any filtering process (such as a profanity filter), and material that has some purpose of being offensive (this last one is not because the Logicracy staff has any respect for political correctness, but it exists simply to avoid what almost always turns into a flame war). This does not mean you should deny or hide well-known or proven facts that certain groups may disagree with. It does mean you should state those facts as you would to a friend you care for, and only if you feel it is important to do so.

Forums, by their nature, invite argument. Arguments must be civil. They should be debates rather than arguments. They should involve exchanging facts, not flames. Do not place wording in a post that has even a partial intent of angering another poster. Certainly some posts will make others angry. Just don't go out of your way to add unnecessary wording that causes anger.

Except for the forums in the Logicracy category, posting to forums will cost you forum points (FP). Your forum points are kept separately for each subject. When you post to a forum, you will be charged FP according to the number of characters in the post, The FP is taken from your FP accounts in the subjects associated that forum. For example, the "Earth sciences" forum will charge FP from the oceanography, geology, meteorology, and geophysics subjects. It is not necessary to have FP in each of those subjects to post. The FP you are charged from each subject is proportional to the FP you have in that subject and calculated so the total FP taken is proportional to the number of characters in the post times a forum character price.

What are moderators?

Each forum has at least one moderator that has absolute power over the posts and threads in the forum (within the dictates of the terms). Moderators can freeze threads, delete threads, delete portions of a post or entire posts, edit posts, etc. Moderators may post messages or send private messages dictating what is acceptable and what is not acceptable in the forum. Moderators can also alert the Logicracy staff of the behavior of certain users. If you disobey a moderator who is obeying the terms, then you are subject to disciplinary action.

The moderator must, of course, obey the terms , as well. If you observe a moderator disobeying the terms, please consider reporting them. You may also civilly post in the forums your disagreement with any part of the terms, site design, particular rules being enforced in a forum, and moderator behavior in a given forum, as long as you obey the terms and forum rules.

There are, of course, moderator rules that are provided privately to the moderators. The moderator rules are meant to make interaction with Logicracy a pleasant and useful experience for the most people possible. Neither the rules nor the moderators will be perfect. From time to time moderators will be too severe or too lenient, or make blatant mistakes. It is human nature. We ask that you forgive them for this (but consider reporting them, of course, if they are violating of the terms). In order to avoid public debate, the moderator rules are not public knowledge, except for the following rules:

1. Moderators are discouraged from commenting on their opinion (good or bad) of the actions of Logicracy staff, terms, the forum rules, the actions of other moderators, or legal matters involving Logicracy.
2. Moderators are not allowed to respond to the user sending in an abuse report. (This reduces the temptation to discuss any action or inaction taken for the report).

Moderators are chosen by Logicracy staff. One of the criteria for choosing a moderator is that there be no known evidence that the user has an ambition to be a moderator. Therefore asking to be a moderator or being didactic in the forums is usually a sure-fire way to blow your chances of becoming a moderator. You are encouraged to point out the rules when a rule is violated or being discussed in a discussion in which you are involved, just don't be didactic or go out of your way to find and respond to violations.

If you wish to report an offence, read the section on reporting abuse .

Why can't I use even mild profanity?

• Profanity is almost exclusively used to exaggerate or convey emotion, which hinders the purpose of Logicracy.
• Logicracy will work best when it is not perceived to be ignoble.
• If there is any correlation at all between intelligence and use of profanity, it is unlikely that it is generally a positive correlation.
• We know it is possible to easily communicate well without profanity, since great writers of centuries past had no problem doing it. 

Let's say there are two people. All that is known is that one regularly uses profanity and the other does not. Which one would most people pick to decide whether the nation should go to war or not?

What is expertise level?

Your expertise level in a given subject, or 'level' for short, is an integer indicating a combination of your skill rating and overall experience in that subject. Your top several levels are displayed in your forum posts and on your profile page. Level is not to be confused with skill rating . Level is for show, only. Skill rating is recorded in the raw data, and used in analysis.  

How do I submit a question?

You can submit a question by clicking on the link on your home page. Select the subjects to which you want to submit the question and fill in the form fields. Then click the Preview button. From the preview page, press your browser's BACK button to return to the the question edit page from where you can again edit the question. Once you are sure the question is the way you want, press the 'Submit' button.

When press the 'Preview' button, you will be told whether you have enough question points (QP) to submit the question. Your QP is kept separately for each subject and will be charged from each subject you specified in the subject list for the question. You earn QP by answering questions well in that subject. Your current QP for the currently selected skill-testing subject is displayed on your home page. The subject list that is displayed when you go to the "Select subjects" page shows how much QP is charged per question text character.

A question contains an optional body text and from one to six optional text answers.

You can also specify the URL of an image to display directly with the question . This image URL must point to an image on another server since Logicracy does not serve image files. Images may be GIF or JPG files.

There are several question styles that can be created using the available question fields. For examples:

1. The body text can be a textual question and the optional answers are just that--optional answers to that question.
2. There is no body text, and there is only one optional 'answer' which is simply a statement that the user rates to indicate how much he agrees with it.
3. There is no body text, and there are several optional 'answers' which are mostly mutually exclusive statements, and the user rates each to indicate how much he agrees with it.
4. An image is displayed that contains several numbered sub-images. The body text asks something about the sub-images and the optional answers are simply numbers (or letters) corresponding to the sub-images. The user then rates each sub-image by rating the corresponding optional answer.
5. You can specify in the body text or an optional answer text that the user should use the radio buttons of a single optional answer to specify a numeric range. For example, the body text, "How far apart are the Sun and the Earth, in millions of miles?" might be followed by a single optional answer with text, "A=less than 10, B=10 to 50, C=50 to 150, D=150 to 500, E=greater than 500". However, you should use this technique only when you think an answerer should be partly rewarded according to how close his answer is to the correct answer. Also, consider that the ranges you choose may give the answerer a hint at the right answer.

When you submit a question, you do not tell the system which of your optional answers are correct, nor do you tell it whether the question is opinion or skill-testing. That will be discovered by the system from the way users are answering it.

See How should I judge a question for more details of question submission rules.

How should I judge a question?
What types of questions should I submit?

You should judge a question according to the following criteria. If you want a question you submit to be judged well, then...

1. Questions should not violate the site Terms . Prohibited material includes, but is not limited to, advertisements, private information and property, lies (except when used as an optional answer, and only if it doesn't violate the other rules), plagiarism (except from other questions), slander, material inappropriate for minors, abusive or obscene material, gibberish or obfuscated material, improperly attributed material, etc.
2. Exact plagiarism of existing questions is, of course, not allowed. However, you are encouraged to make even minor changes to the text of an existing question and re-submit it--even if it asks the same thing. This is so that the question turn-over rate is increased, which helps deter bot developers from compiling a list of questions and their right answers (since a bot must have pretty much the exact question text in order to work). Ideally you should make something more than a minor change to an existing question, or simply compose a new question, so that there is more variety among questions and in case someone develops a bot that can recognize and understand minor changes in a question.
3. Do not create an optional answer that is a catch-all, like "all of the above" or "none of the above". Doing so could cause some users to presume they don't need to individually rate each answer while other users still do.
4. Do not require your optional answers to be presented in some order. In the future the order might be randomized by the system. Currently, optional answers are placed in reverse alphabetic order. 
5. After reading and understanding a skill-testing question, a person of moderately high skill in the subject ideally should have a 2/3 probability of correctly answering the question within twenty seconds without any other tools than his mind. Answering the question should not require a calculator, technical reference, pencil and paper, etc.. (Although users are welcome to use such tools if they want)
6. Do not rate the quality of an opinion, demographic, or personal question poorly only because you think it is trivial or a mish-mash of issues. Analysis of the answers to trivial questions often reveals interesting correlations.
7. Skill-testing questions should not be a test of obvious or unimportant trivia or some esoteric knowledge or skill that notably exceeds what most would consider evidence of high skill in the question's subjects. For example, a question in the 'mathematics' subject should not test the user's knowledge of digits of the square root of two--that's what computers and calculators are for. A question in a history subject should not test a user's ability to know who made some little-known statement, regardless of its profundity.
8. Try to avoid submitting very many questions that have only one right answer and several wrong answers. Quiz takers can take advantage of this fact by answering all optional answers with a "disagree" when they are not sure of the answer. The consolation, however, is that such a question will be automatically rated as easy, and this won't yield as much skill rating when answered well.
9. Questions should be a good test of knowledge, intelligence, skill, predictive ability, competence, character, opinion, emotional response, demographics (personal information), and/or consistency (in opinion) in at least one subject.
10. Questions should contain no spelling errors (unless the purpose of the question merits it).
11. Questions should contain proper United States English grammar and spelling. For example, use "color" not "colour". Use measurement units used in the United States: Use "miles" not "kilometers"--(unless the purpose of the question merits it). You can, if you like, include a metric measurement in parentheses following a U.S. measure. For example, the moon is about 240,000 miles (386000 kilometers) from the Earth. A "million" is a thousand thousand and a "billion" is a thousand million. Fractional values should use a period (.) to separate the fraction from the integer part, and commas may be used to separate sets of three non-fractional digits starting from the decimal point (ninety-three million miles can be written 93,000,000 miles)--do not use these symbols any other way in numeric values.
12. A single optional answer should be a single statement, not a set of answers--which is a form of spin. Otherwise there is a danger that a user will agree with one of the statements but disagree with the other statement in the same optional answer. This would taint statistics because different users with the same opinion might decide to answer it differently.
13. Questions with images should comply with the image submission guidelines . Please rate a question poorly if you encounter a question that does not.
14. Questions should have no missing or repeated optional answers, formatting errors, etc.. Optional answers must be consecutive starting with answer number 1. The Body text is optional, as is the image URL.
15. Skill-testing questions must not have unrelated optional answers combined into one question. For example, you should not use each optional answer as a separate and unrelated statement to be rated by the user. This is because the quality and subjectivity ratings of the question can apply only to one issue. Opinion and demographic questions may have unrelated optional answers, however. In fact, doing that makes analysis easier and more reliable.
16. Questions should contain no spin. Questions designed to test emotional response should not, however, be judged poorly. See the section on spin for a discussion of this.

What should I know about submitting images?

Why can't I submit a sound under the "Fair use" doctrine?

What is spin?

Whether a question has spin or not is part of the quality judgment you make when you answer a question. In the end, the definition of 'spin' is up to you. However, here are some suggestions:

A question containing spin includes words or statements that will likely be interpreted differently by different users, prompts for a single answer for multiple questions, presents a false or highly debatable statement as truth, mixes conflicting definitions of the same word or phrase, or in your opinion fails to mention a critical fact. Optional answers may, of course, contain false statements, but a single optional answer should not contain more than one statement whose truth should be rated separately.

Anything other than the above, is not spin.

Do not judge a question poorly simply because it is complicated or rather difficult (as long as a user fairly skillful in the subject could probably answer it within 20 seconds without extra tools), requires one to carefully note the wording or graphics in order to correctly understand the question (as long as there is no obfuscation), or requires understanding of a reasonable prerequisite concept in order to understand the question. For example, a graph with truncated axes (as long as they are clearly marked) should not be considered a "trick" question--ability to recognize truncated axes is a reasonable prerequisite of good skill.

Do not rate poorly a question that tests for emotional response. Testing for emotional responses is not spin. Submitting such questions is encouraged. For examples:

1. The user is asked whether he thinks a despised politician is smart enough to do something laughably simple, like add two single-digit numbers together. Users that don't take Logicracy seriously and/or prefer to be ridiculous rather than logical will answer that the politician is not smart enough to add the two numbers when in fact, we all know he probably could. Joking around has its place, but not in Logicracy.
2. A question of the form, "Would it be a good thing if Joe Politician pushed for law X", is asked in two separate questions, once when 'Joe Politician' is replaced with the name of a beloved politician and once when it is replaced with the name of a despised politician. All else being equal, it seems an emotionally controlled user is more likely to answer the questions differently, and analysis could reveal this.
3. A question is asked while an image of something unrelated is displayed--like that of a politician. The question need not be related to the image. It seems emotionally controlled users will be affected by such things, and the effect will appear in analysis.

The best questions are mathematically precise--almost written in legal-ese. Being verbose (without ambiguity) is the best preventative medicine for a bad case of vagueness. So do not condemn a question or its optional answers for being a little verbose. Verbosity is a sign of good legal-ese! Rather, condemn a question for being vague or ambiguous.

For an example of spin, the question "Is <X> wrong if no harm is done" mixes conflicting definitions of the word "wrong". The question forces a definition of "wrong" that does not include the concept of causing harm, yet a far more popular definition of "wrong" is "that which causes harm". Consider: "Does thievery cause harm as long as it doesn't cause harm?". Many are forced to answer that <X> in this case is not "wrong", even though they may believe <X> practically always causes harm. Also, the question forces the voter to imply that practical laws could be created and reliably enforced to cover rare exceptions (when harm really isn't caused by thievery).

For another example: Users are asked how much they agree with the statement, "Since global warming is really occurring, we should implement the Kyoto protocol". This question causes the reader's single response to be a mixture of five separate unrelated opinions, making it easy to intentionally misinterpret the results of the question. The opinions are: Is the temperature of the surface of the Earth really increasing, is global warming really all that bad, is global warming caused mostly by man, can politicians world-wide most likely implement the Kyoto protocol without notable corruption, and is the Kyoto agreement a good thing if it could be implemented efficiently. Also, some users might consider that certain critical facts are missing, like the fact that if the Kyoto agreement were fully implemented by all nations on the Earth, greenhouse gas emission produced by mankind would only be reduced by about three percent; or the fact that the Martian ice caps are melting notably faster than the Earth's. Before submitting a question on a controversial issue, try to view it with the eyes of your adversary, and address every valid complaint he might have.

For another example, the question "Does war cause more deaths or less deaths?" neglects specifying what alternate to "war" is being referred to. The term "war" here can mean "starting a war" or it can mean "ending a war started by others". Even "Does starting a war cause more deaths or less deaths?" is vague because one nation may start a war to stop a form of killing not called "war", like genocide or unquestionable threat of war.

Also, questions should address probable scenarios. Asking for an opinion on an extremely improbable scenario may not really be spin, but it is a waste of resources.

Finally, "trick" questions, whether intentional or unintentional, should be judged as containing spin. Common English communication takes advantage of certain presumptions for the sake of brevity. If the question is asked, "What is the shortest distance between any United States land and any Russian land", the answerer will be alerted to the specific meaning because of the presence of the words "any" and "land". If the question were asked, "What is the distance between the United States and Russia", a typical answerer may very well presume that the question refers to common routes of travel between more populated areas, because that is precisely why such a question would be asked in casual conversation. Answering the second question with "2.5 miles" would be technically correct, but probably not the answer that was really sought and such an answer would certainly be considered presumptuous in casual conversation. (Diomede Island off the west coast of Alaska is 2.5 miles from Russia's Big Diomede Island). 

How do I get raw data and produce and display my own analyses?

Overview

Every time a user answers an opinion or demographic question, everything about that user's answer is recorded. Each record of information represents a single answer of a single question by a single user. A record includes things like question number, user number, user's answers, user's skill for each subject, and more. These records are sometimes referred to herein as "raw data". Since the Logicracy server has limited storage space, older records are eventually deleted. Typically up to several months of data are available.

You can search for, purchase, and analyze records from this database by spending the AP (analysis points) that you have previously earned by answering questions well.

You'll need to know the question numbers of the questions you are interested in. For this reason you should remember the question number you see on the quiz page when you encounter a question that you would like to include in an analysis. You can also click on the 'Search for questions' link on your home page to search for opinion and demographic questions containing certain text. When you search for a question on that page you will be charged AP.

You'll also need to know the subject numbers of the subjects you are interested in. Click on the 'Change current subjects' link to see all the subject numbers. The subject numbers are listed in the first column of the subject table.

Once you know which question numbers and subject numbers you want to use in your data gathering and analysis, click on the "Perform an analysis" link on your home page. A page will display that allows you to specify analysis parameters.

First, appropriately check or clear the checkboxes that let you specify whether you want raw data to be included in the output, and whether you want a scatter plot created. Each of these costs AP.

Then specify the maximum number of records to retrieve. You would do this to conserve your AP.

Then specify the data set by entering an expression into the "Filter expression" box. The format is described in detail below.

Finally, if you specified that you want a scatter plot created, fill in all following boxes with the expressions describing each attribute of the scatter plot (described below).

Once you have specified your analysis parameters, click the "Submit" button at the bottom of the page. You will either be alerted of a syntax error in one or more of your expressions (in that case, click your browser's Back button to return to the analysis page) or an analysis result page will be displayed. Do not close the analysis result page until you have saved it! Otherwise you'll have to spend more AP creating another analysis. When the analysis result page is displayed, you can use your browser's "save page as" function to save the analysis page locally. If the result includes raw data, you can copy and paste it to a text file to remove the HTML formatting characters. If the analysis included a scatter plot, you can right-click it and save the image locally.

The rest of this section describes in detail the analysis parameters you can specify on the "Analyze" page, strategies in using them, and how AP is charged.

Filter expressions and Max records:

Enter into the "Filter expression" box the criteria to limit and control the records that will be retrieved. The Filter box defines your data set. A typical analysis might include one question number specification. For example, the following filter text will retrieve answers of question number 1608:

question_num=1608

For another example, the following filter text retrieves records for questions 31006 and for question 119, but only those in which the answering user's skill in subject number 18 is greater than .5.

(question_num=31006 OR question_num=119) AND skillx18 > 128

You must also specify a numeric value in the 'Maximum number of records' box. No more than this many records will be retrieved.

If the number of records actually retrieved hits the limit specified by the maximum records limit or exhausts your AP, then the data set retrieved is neither random nor complete. In that case the record set that is retrieved depends on the internal workings of the database engine, which depends on a variety of non-random factors. The RAND keyword (described below) can be used to improve this situation. For example, the following filter text will retrieve only about ten percent of the questions that match the criteria, which, hopefully, is less than the maximum records limit you specified and less than the maximum records you can afford according to your AP:

question_num=31006 AND skillx18 > 128 AND RAND() < .1

You are not given a chance to experiment with your filter criteria without being charged AP.

If you are retrieving raw data so that you can perform analysis locally, then you might be able to save AP by getting small non-intersecting sets of data until you have what you need. For example, if you want approximately the top 500 records for question number 43 in which the user has the highest skill in subject 'music', then first try a filter that you believe will get far less than 500 records by specifying that the music skill must be far higher than you think is needed. If that criteria returns less than 500 records, then get more data but specify the next lower range of music skill in your criteria. Continue to get data with lower and lower ranges of music skill value until you have enough records. Then concatenate the saved data files.

In the 'data filters' edit box you can enter one or more arithmetic expressions separated with AND or OR. You can also group expressions with parentheses. Standard arithmetic operators and syntax apply.

For those readers familiar with MySQL, the filter text is a WHERE clause, but with restrictions to prevent hacking. For those unfamiliar with the syntax, the following table of examples might help (see the table at the end of this section for a list of variables that may be used):

scatter plot expressions:

If you checked the checkbox that indicates you want a scatter plot created, then you need to specify the scatter plot parameter expressions.

Each of the six boxes defines a chart parameter. Place in each box an arithmetic expression using standard math syntax and functions (MySQL syntax), and any of the variables in the below list. Some typical expressions are pre-loaded in the boxes, as examples. [editors note: more information and examples will be added here soon].

The following is a list of variables that may be used in any of the filter or scatter plot expressions:

How analysis points are charged:

(Note: AP is not yet charged. This text represents future behavior of the site)

For every analysis you perform, you will be charged analysis points (AP) according to the following equation:

AP = NumberOfRawDataRowsListed * ChargePerRawDataRow  +  NumberOfPointsInPlot * ChargePerPlotPoint

Where:
AP is the analysis points that you are charged.
NumberOfRawDataRowsListed is the number of raw data records listed. If you didn't specify raw data be included, then this is zero.
ChargePerRawDataRow is the charge, in AP, for each record of raw data that is listed. This value is not yet defined.
NumberOfPointsPlot is the number of points in the scatter plot. If you didn't specify that you wanted a scatter plot, then this is zero.
ChargePerPlotPoint is the charge, in AP, for each point in the scatter plot. This value is not yet defined.

If you didn't have enough AP to perform the analysis you requested, only partial data will be displayed and you will be told so on the analysis page. You are still charged AP for this. It is important to realize that AP will be charged no matter how well or how poorly you chose your analysis parameters. You will likely need to spend a considerable amount of analysis points simply learning the process. You should study this section well before attempting an analysis. While experimenting, you will likely want to limit the total number of raw data records you retrieve to something quite small, like 10 or 20. See the LIMIT keyword described below.

What is my public profile?

When you create an account in Logicracy, you are given a single web page that you can somewhat customize and that you can tell others about so that they can see your statistics in Logicracy, personal comments, a picture of your face, etc. Profiles are available via the forums. To view a user's profile, go to the member list in the forums and click on the user name. You can edit your own profile by clicking the "Profile" link on the main forum page.

How can I prove my skill in Logicracy in my other writings on the Internet (blogs, my web pages, forum posts, etc)?

Tell me about the image captchas. Will I be penalized for missing a few of them?

An image captcha (distorted image of some text that the user must resolve) is given with each question. The image captchas used by Logicracy are very effective in deterring bots because...

• They don't require the user to type anything--the user need only select one of three buttons to answer the captcha.
• They require the user to have read the question. (Although it is true a client-side bot might be created which would allow a human to type the word he sees and then the bot could search for it within the question text--but that is notably more work on the part of the human).
• Since they require no typing, they can be given far more frequently--making the test more reliable and making it completely inopportune to hire humans to answer them, especially since the humans must have either read the question or must type in the word they see.
• Since a reliable success ratio can be measured because the tests are given more frequently, they can be made a little more difficult to recognize because it is OK for humans to miss a few. Mainly this means that letters can be allowed to overlap. This takes advantage of the fact that humans can recognize words with overlapping letters far more often than can bots.

Don't worry about missing a few of the image captchas. Everybody will. Some of them are just too distorted to read. The tests are adjusted so that humans still miss some and bots miss many more. The system does not penalize you unless your captcha success rate is notably closer to that of a bot than a human.

What browsers have been tested with Logicracy.com?

Mozilla Firefox version 1.5.0.9
Opera version 9.10 build 8679
Microsoft Internet Explorer