The world hasn't had a good method for deciding truth and establishing it in a transparently correct way. TruthSift solves this problem.

Topics may be browsed in different *layouts*. Clicking on the Gold, Red, or Blue Box on the home page will take you to a view centered, respectively on either the Topic, most recent *Con*, or most recent *Pro* statement. You can walk around to see the rest of the topic by double click on other statements or selecting them and choosing center. A focused view is recommended for large topics. Alternatively, Clicking on the Topic Title on the home page will display the whole graph.

TruthSift rates a statement "*Tentatively Established*" (usually abbreviated simply as *"Established"*) only when there is
a proof of it with no step in active dispute. A demonstration that
begins from observations that are unchallenged and proceeds through
unchallenged proofs with no step of the demonstration in dispute.

If anybody disputes any component of your proof, and can state a reason why its in error, they can add a challenge to the component, and if they want a supporting network of proofs and assumptions for their point of view. (EXAMPLE 2)

TruthSift asks of its members that you believe every public post you make is correct, not duplicative of a parallel post in the topic, and clearly stated.

By correct we mean, does rationally prove or refute its conclusion. Ad hominem attacks are not permitted. The goal should be to create a clear exposition of what can be proved, and how the principal challenges fail. Think of yourself as publishing a paper in the scientific literature.

Don't get hung up on whether a statement should be added as a proof or an assumption of another until the matter is challenged. Frequently you want to assemble arguments for a proposition stating something like "the preponderance of the evidence indicates X", and these arguments are not individually necessary for X, nor are they individually proofs of X. It is safe to simply add them as proofs of the above proposition.They are not necessary assumptions, and if not enough of them are established, the target may be challenged on that basis.The goal is a diagram that transparently explains a proof and what is wrong with all the objections people have found plausible, with noone finding more rational objections. Edits that move in that direction are useful and desired.

Probability Mode may be toggled on or off.

Also Citations Mode may be toggled on or off. In citations mode a statement will not be considered established unless it is a citation or has an established proof (and would also be established in normal mode.)

If either of those changes are made, the topic may be "Saved-As" by clicking on a statement and selecting save-as. The Probability Mode status and Citations Mode status will be saved with the new diagram copy.

Un-Saved Mode allows one to edit locally without the changes being updated to the public topic. You will have to turn it back off again if you want to save.

Statement tip toggles on and off the q-tip that is by default displayed when you mouse over statements.

Probability Mode is turned on for some topics, as discussed in the section above. If it is on, the depth of shade of the statements indicates the probability attributed to the statement. Statements are still pink or blue depending on whether the are CON or PRO, but the darkness of the shade indicates their probability. Also the specific numerical probability assigned to the statement may be viewed by hovering the pointer over the statement, and is shown at the bottom of the q-tip.

TruthSift Probability mode supports the easy collaborative construction of powerful probabilistic models related to Bayesian Networks . Each statement has a parameter called proposed probability(PP) which is 1 by default, but may be set between 0 and 1. Scroll down in the edit window for the statement to see the setting. The proposed probability sets the likelihood with which the statement claims something happens. If the statement is established, then this something will happen with likelihood PP. For example, the statement may be a proof, in which case it will establish the proof with likelihood PP. Or it may be a challenge, in which case it will establish the refutation with likelihood PP. A proof can be used, for example to model a statement that will sometimes cause another statement to become true, or a challenge to sometimes prevent its being true.

Test statements claim that some observation provides evidence some other target statement is more likely True or more likely False. Test statements have two additional parameters: Likelihood if Target True, and Likelihood if Target False. The likelihood if Target True estimates the likelihood of some observation given that the target of the Test connector is True. The Likelihood if Target False estimates the likelihood of the observations if the target statement is false. These parameters are set by members, (who may add proof statements to justify them). TruthSift then assigns the Test a Likelihood Estimate (LE) = Likelihood if True/(Likelihood if True + Likelihood if False)

For example, an epidemiological test may report results that would have likelihood 95% of happening if the test is unbiased and the target statement is true, 5% of happening if the test is unbiased and the target statement is false. If the test is hopelessly confounded, the test results tell us nothing about the statement. The Probability of the Test Statement itself estimates the likelihood the Test statement is an accurate test, and when it is an accurate test, it estimates the target statement has a Likelihood Estimate of .95=.95/(.95+.5)

TruthSift then displays a Monte Carlo estimate of the probability each statement in the whole topic is true, marginalized over all the above parameters. Instances are generated uniformly from the distribution implied by the model, and weighted according to the Tests. The probability of a statement is given as the weighted likelihood it is true.

Instances are generated starting from the leaves and assigning a 1 or a 0 to each statement in turn once its parents have all been assigned. If a statement has all its assumptions from 1 statements, no 1 challenges, and if it has proofs, at least one must be from a 1 statement, then it is assigned a 1 with probability its proposed probability. Otherwise a statement is assigned a 0. After an instance has assigned a 0 or a 1 at each statement in the topic, the instance is assigned a weight by the product over test connectors, of 1/2 if the Test statement has a 0, of the Likelihood Estimate(LE) if the Test statement has a 1 and the Target a 1, and of (1-LE) if the Test statement has a 1 and the Target a 0. This weights the instances exactly according to the probabilities asserted by the true Tests, and treats the false tests as uninformed about their target. Thus the Monte Carlo samples directly from the stated distribution.

This model allows members to create directed acyclic graphs of possible causes (that act independently with a proposed probability to cause or imply a statement to be True), challenges, that act independently (with a proposed probability) to cause a statement to be False, overruling any proofs of it, necessary factors, that will make a statement false if they are not true, and tests, that assert some hypotheses are favored over others by some observation. TruthSift samples exactly the distribution specified by the model, and reports the likelihood each statement is true. These topics could be used for collaborative, verified, risk models; to support proofs with additional confidence tests; to reason about hidden causes; or many other novel applications.

All members may post and edit public topics. Premium membership (automatically included now, later free for the first 3 months, $2/month thereafter) allows a user to host and participate in private topics for associates or colleagues. We hope you will find this useful for your work, planning, and studying. You may invite individuals or groups using the *my sharing* link below the topic listing on your *my private topics*, or *my public topics* page. Friends may be invited either by supplying an email address or a TruthSift username. They will receive an email invitation to participate with a link. (If they are not yet members, they will have to sign up to edit.) You may specify either *View Only* or *Edit* access (for Private topics). If you want to repeatedly invite a group, you may create groups on your *My Groups* page. Individuals who
haven’t been invited to a private topic will be unable to see it.

You may sign up for different levels of notification on your *Profile and Settings* page. It is useful to be notified when somebody edits one of your statements, or challenges it. You may also upload an image that will be displayed on your user page.

We are in beta and users with questions or comments or suggestions for features or improvements you’d like to see are kindly requested to email them Here(info@truthsift.com) or post or participate in a public topic arguing for them.

More features coming soon.