Claims are Still Not Evidence

 When I read the statement "Claims are not evidence", it almost seems like a truism. It appears so self evident that it's not even worth analysis. However, some people want to dismiss this statement as "demonstrably" false. So in this post, we will analyze the claim and decide whether it is conceptually meaningful. The statement can be summarized as follows: "The fact that someone states X does not provide evidence for the truth of X". In other words, the mere fact of asserting X provides no justification for X being true.

If you are critical, an immediate counter-example can be thought of. Suppose your friend tells you they bought a soccer ball. There are two scenarios; they either bought it or did not buy it. You reason that people don't normally deliberately lie about seemingly trivial claims, therefore finding it rational to believe their claim. The fact of them saying it should count as evidence in favor of the belief they purchased the ball; simple enough right? The problem with this is illuminated when you ask for justification. On its own, the claim is a truth-bearer; the claim acts as a medium containing an assertion about reality, which requires justification if someone deems necessary. In this instance, the claim itself is supported by general background knowledge, auxiliary assumptions (How to never be wrong), and prior beliefs, guided by common sense reasoning and defeasible generalizations about how things typically work in a communicative situation, and subject to conversational Gricean principles. You simply take for granted their assertion due to the plethora of information assumed to simplify reasoning in these sorts of trivial situations; it is not that the claim is evidence, it is that your reasoning is guided by heuristics that simplify the process of determining what utterances to accept. Alter the context ever so slightly, and you may need to ask for justification. On its own, the claim is merely an assertion; it is essentially the conclusion of a chain of reasoning that is suppressed for pragmatic purposes. The act of stating a proposition does not make it true. In the example above, your friend is reporting something in the form of an utterance. Most information needs to be communicated through utterances and not all utterances express propositions; Speech acts are performative and serve multiple functions other than expressing propositions. Your friends report might direct your attention to the facts and evidence that would make their claim true or false; perhaps by consulting a store receipt in this scenario.

Lets consider another example of an utterance which claims something about reality, in order to demonstrate how the mere act of uttering a sentence is not evidence for the truth of the proposition contained in the sentence. Lets assume that the utterance is performed in a context of speakers who share a set of background assumptions. Your friend now tells you "Measles are caused by eating too many rotten eggs"; now they are not simply reporting one of their actions, but they are making a causal generalization within the empirical domain which conflicts with background assumptions. The claim might be surprising to you, after all it does communicate information; in an information theoretic context, a report of the known cause would carry very little information but maybe your friend is so convincing that the surprise contained in this piece of information prompts you to reevaluate your belief about the statistical regularity, causal factors, and prior information related to the topic. The problem is that information is not equivalent to evidence; information becomes evidence after a process of establishing its credentials (through a chain of reasoning); you need to demonstrate that the information acts as a force in favor of a proposition. You can think about it from a statistical modeling perspective; data fed into a model essentially goes through an algorithmic transformation process in which the output is the evidence needed for a decision or to support a hypothesis; if you use garbage data you will have garbage inference. The data becomes connected to the hypothesis through the input-output process of the model, and the research demonstrates the relevance of the output by means of theory or argumentation. Furthermore, the data is most likely contained in a report; but surely we wouldn't call the report the evidence, rather the evidence is contained within the report, which is solely acting as a vessel to transmit the evidence. So in the above example; the fact that your friend asserted a causal generalization is not evidence that the generalization is valid because there hasn't been a process that has demonstrated the statistical regularity. Your friend might be reporting something they believe, but an assertion of a belief does not make it true; you need to have reasons and evidence acting as support. This is done by acquiring facts about the subject matter contained in the utterance and establishing the relationship through some sort of accepted methodology. This is what makes the utterance true or false; not the mere speech act. Your friend is essentially asserting a hypothesis as true without justification. If your friend expresses this utterance and follows with "because X,Y, and Z", then these reasons are support for their claim, and they can be assessed in accordance with evidential standards. Without the evidence, they are merely begging the question. Suppose we have multiple people claiming this same generalization, multiple claims can act as a pointer, directing an evidence collector to investigate whether the empirical relationship holds; but multiple claims are irrelevant with respect to the truth of the causal generalization. Quoting David Schum from The Evidential Foundations of Probabilistic Reasoning, "So we might simply state: A datum becomes evidence in a particular inference when it's relevance to this inference has been established. Unfortunately, no datum comes to us with already-established relevance 'credentials'. The relevance of any datum has to be established by cogent argument." (p.20); in the situation above, a single claim can never justify a regularity.

I've kind of alluded to this before but the main question underlying the phrase is "what counts as evidence"; what belongs to the category we call "evidence"? How can be distinguish between evidence and non-evidence? A formal definition of "Evidence" has alluded thinkers for centuries, but it seems that something is evidence with respect to a proposition or hypothesis under examination; evidence doesn't just exist in some Platonic category, as mentioned before we have to establish its credentials. So far I have been arguing that assertions do not count as evidence for the thing being asserted; in fact this is the very reason we need evidence, to resolve assertions that might be under dispute, to distinguish between the likelihood of competing hypotheses, and to ground assertions on facts. David Schum describes how the concept of evidence has evolved over the ages; (p. 13) "Hacking argues that before the 1600's the concept of evidence involved either testimony or authority and that both involve people and not things. Testimony comes from human witnesses; authority was provided by records of ancient learning on part of renowned thinkers. For example, early court trials in England did involve the sworn testimony of witnesses who, unfortunately, had another characteristic; they were frequently jurors as well as witnesses in the same trial. Medical diagnoses were based upon the recorded advice of ancient scholars such as Hippocrates and Galen. Arguments about other matters were often settled by reference to the teachings of persons such as Aristotle and St. Augustine. What was missing until the 1600's, says Hacking, was a concept of evidence that included physical objects or things that could point or serve as signs for other things.", in other words, evidence became empirical around the Enlightenment. This is typically what we consider as "evidence" in the modern era. But notice that above, I am not disputing the merits of using evidence from testimony (see epistemological problems of Testimony) or authority. The crucial difference however is that testimony and authority are by definition something more than merely a claim to truth; testimony presupposes the person testifying was a witness (therefore having access to evidence), they are not just asserting a claim. Of course evidence does not always involve that which we can observe in some way; but in the case of testimony we typically take it to mean that the person testifying was in some position to know, observe, or record some sort of facts that we do not have direct access to. These forms of evidence become surrogates for the observations we are unable to make. In the case of authority, they are not merely asserting something as true; they provide justification based on methods that are typically reproducible. In other words, we don't just accept their claims, we ask for justification which they can provide, almost always with reference to some body of evidence. With respect to expert opinion (authority), we accept their claim for pragmatic purposes, because we cannot synthesize all of the information they have with respect to the topic. Again, in this case, they are acting as a surrogate with respect to the actual evidence used to support their conclusions. Below is the Scheme for Argument from Expert Opinion from Douglas Walton's Argument from Expert Opinion as Legal Evidence: Critical Questions and Admissibility Criteria of Expert Testimony in the American Legal System:

• Major Premise: Source E is an expert in subject domain S containing
• proposition A.
• Minor Premise: E asserts that proposition A is true (false)
• Conclusion: A is true (false).

The claim is the conclusion of the syllogism; it should be obvious that this is not the evidence. When we ask critical questions, we can probe the actual evidence buttressing the claim. The questions below focus on whether their claim is relevant; in other words, answering these questions helps demonstrate that their claim can act as a surrogate (most likely during a cross examination). Obviously if their assertion is merely a claim (fails on number 6), we would discard it:

1. Expertise Question: How credible is E as an expert source?
   1.1 What is E’s name, job or official capacity, location, and employer?
   1.2 What degrees, professional qualifications or certification by licensing
   agencies does E hold?
   1.3 Can testimony of peer experts in the same field be given to support
   E’s competence?
   1.4 What is E’s record of experience, or other indications of practiced
   skill in S?
   1.5 What is E’s record of peer-reviewed publications or contributions to
   knowledge in S?
2. Field Question: Is E an expert in the field that A is in?
   2.1 Is the field of expertise cited in the appeal a genuine area of
   knowledge, or area of technical skill that supports a claim to
   knowledge?
   2.2 If E is an expert in a field closely related to the field cited in the
   appeal, how close is the relationship between the expertise in the
   two fields?
   2.3 Is the issue one where expert knowledge in any field is directly
   relevant to deciding the issue?
   2.4 Is the field of expertise cited an area where there are changes in
   techniques or rapid developments in new knowledge, and if so, is
   the expert up-to-date in these developments?
3. Opinion Question: What did E assert that implies A?
   3.1 Was E quoted in asserting A? Was a reference to the source of the
   quote given, and can it be verified that E actually said A?
   3.2 If E did not say A exactly, then what did E assert, and how was A
   inferred?
   3.3 If the inference to A was based on more than one premise, could one
   premise have come from E and the other from a different expert? If
   so, is there evidence of disagreement between what the two experts
   (separately) asserted?
   3.4 Is what E asserted clear? If not, was the process of interpretation of
   what E said by the respondent who used E’s opinion justified? Are
   other interpretations plausible? Could important qualifications be
   left out?
4. Trustworthiness Question: Is E personally reliable as a source?
   4.1 Is E biased?
   4.2 Is E honest?
   4.3 Is E conscientious?
5. Consistency Question: Is A consistent with what other experts assert?
   5.1 Does A have general acceptance in S?
   5.2 If not, can E explain why not, and give reasons why there is good
   evidence for A?
6. Backup Evidence Question: Is E’s assertion based on evidence?
   6.1 What is the internal evidence the expert used herself to arrive at this
   opinion as her conclusion?
   6.2 If there is external evidence, e.g. physical evidence reported independently of the expert, can the expert deal with this adequately?
   6.3 Can it be shown that the opinion given is not one that is scientifically unverifiable?

As mentioned above, testimony is evidence (albeit, very weak at times, at other times it is a deciding factor) because of the fact that the person testifying is supposed to be a witness. In this case, the person testifying is using a locutionary act to express or demonstrate one of their perceptions. The perception itself (the act of witnessing something) is the evidence which we need to validate; the claim is merely an assertion if they can't fall back on some memory. If they were not in a position to know, we disregard their claim as merely an assertion (no evidence). If they were not a witness, their assertions become hearsay and are immediately disregarded as inadmissible in a court of law. We need to be careful though not to confuse different types of claims in which testimony would be completely irrelevant. Remember from above, if someone asserts a causal generalization about the observable world, testimony will not constitute evidence (we will need scientific evidence). Testimony is useful in circumstances where we are not in a position to know whether an event occurred and when there is no physical or trace evidence from the event. Making a claim from a position to know whether an event occurred can greatly simplify the search for empirical evidence in some circumstances, but is completely irrelevant when it comes to statements concerning diagnosis, generalization, prediction, evaluation, and many others. Consider the statement from someone "in a position to know"; they claim that they have clinical depression.

• Major Premise: Source s is in position to know about things in a certain subject domain f containing proposition p.
• Minor Premise: s asserts that p is true (false).
• Conclusion: p is true (false).
• CQ1: Is s in position to know whether p is true (false)?
• CQ2: Is s an honest (trustworthy, reliable) source?
• CQ3: Did s assert that p is true (false)?

The fact that they are "in a position to know" since it is literally about them, does not act as evidence for the clinical diagnosis; you need someone with the proper evaluation criteria to make that judgement. And again, the fact that someone claims they have depression, does not mean they have depression. This really should be clear by now. The clinician will make the assessment, making some form of argument from authority based on their credentials. But this will always be with respect to observable evidence and the diagnostic criteria (THE METHOD) for discerning the truth of the proposition. Maybe you see another clinician as a source of corroboration; they use an alternative method of diagnosis that is also robust.

When I think back to what a claim even is, I recall my first debate course in university. The "claim" was the thing you needed to defend during an argument. There are claims of fact, in which your interlocutor asserts something true or not, claims of value, in which you assert that something is more or less desirable (or of moral goodness), claims of policy (practical reasoning) in which one asserts a course of action is superior to another (perhaps based on a cost benefit analysis), there are claims of cause and effect, and many other types. In each case, these are the things needing justification; they do not stand on their own. We were introduced to the Toulmin model of Argumentation, where there is a clear distinction between different components of an argument. This is so elementary I really don't understand why there would be a dispute.

• Claim: This is the main point or thesis of the argument. This is what the pro-side is attempting to convince you of or trying to prove. If the claim is not directly stated, just ask, “What is the pro-side trying to prove?” In the sample argument, the conclusion, the claim you are attempting to prove is that "Phil’s friends will lead successful lives."
• Grounds: Here is the starting point of your argument that leads to your Claim. This is what you have observed, read or what you believe to exist. In this sample argument, the grounds are that "Phil has several friends who have graduated from college". This is your data; your evidence.
• Warrants: This is the overall logical underpinning of the argument. A general rule that can apply to more than one Grounds. The Warrant can be a universal law of nature, legal principle or statute, rule of thumb, mathematical formula, or just a logical idea that appeals to the person making the argument. Warrants usually begin with words like; all, every, any, anytime, whenever, or are if-then, either-or, statements. The Warrant is a general rule that has no exceptions. Those come later; Warrants are important because they provide the underlying reasons linking the claim and the grounds. You can infer the warrants by asking, "What’s causing the advocate to say the things he/she does?" or "Where’s the advocate coming from?" In our example argument, the warrant is that “All people who graduate from college are successful.” No exceptions. These are obviously the defeasible generalizations in which your argument depends, they glue the evidence to your claim.
• Backing: Backing is the specific data, which is used to justify and support the grounds and warrant. In Toulmin’s original work, he only includes Backing for the Warrant. I am adding Backing to also look at the quality of the original Grounds of the argument. Critical thinkers realize that there must be backing for their statements or they are merely assertions. When clashing with an argument, we need to look at the quality of evidence that supports the initial grounds. In our diagrammed argument, the Backing for the Grounds are the names of the specific friends who graduated from college. The Backing for the warrant comes from an LA Times article that college graduates earn $1,000,000 more in their working lifetime than non-college graduates do. I like to separate the Backing from the Grounds and Backing for the Warrant as they are two different areas that can affect the strength of the argument. Toulmin makes no such distinction.
• Reservations and Rebuttals: They are the “unlesses” to the Warrant. Reservations do not change the wording of the warrant. Reservations do not change the “universality” of the Warrant. But Reservations are exceptions to the warrant. These exceptions weaken the validity of the conclusion because the Grounds may just be one of these exceptions, thus meaning that the Claim is invalid. In our example, your uncle has a Reservation to the Warrant. He states that people who get a college degree will succeed, unless they are lazy. The “unless they are lazy” is the Reservation to the Warrant. Note that a Reservation does not refer to a rejection of the Warrant. In this example "Unless they did not graduate" would not be a Reservation because it implies that the Warrant did not happen. Instead a Reservation is a statement that suggests that even though the Warrant took place, the Claim may not occur. These are the exceptional conditions which are identified through critical questioning.
• Qualifiers: Suggest the degree of validity of the argument. If there is no Qualifier, then the argument is 100% valid. But if a Qualifier exists then the conclusion is less than absolute. With a qualifier, the argument is about probability and possibility, not about certainty. You cannot use superlatives like all, every, absolutely or never, none, and no one. Instead you need to qualify (tone down) your claim with expressions like; most likely, many, probably, some or rarely, few, possibly, etc. In our sample argument, it is argued that Phil’s friends will “most likely” be successful. The “most likely” is the Qualifier.

If we consider argument from a mathematical perspective, claims cannot be justification or evidence. The lexicon in mathematical logic doesn't even distinguish between claim and proposition. I bring this up because sometimes its necessary to make distinctions between concepts like a theorem and an axiom; is there any meaningful conceptual distinction between a proposition and a claim? And are claims equivalent to evidence? According to the Mathematical Association of America, (See Proofs and Mathematical Reasoning and Mathematical Terminology for more)

What different kinds of statements are there (lemma, theorem, proposition, claim, corollary, etc.), and when is each used?

• Proposition: sum from 1 to infinity of 1/n diverges.
• Lemma: If the limit as x goes to infinity of the integral from 1 to x of f(t) dt diverges, then the sum from 1 to infinity of f(n) diverges.
• Claim: The limit as x goes to infinity of the integral from 1 to x of 1/t dt diverges.
• Theorem: the sum from 1 to infinity of 1/n^p converges if p>1 and diverges if p<1.
• Corollary: the sum from 1 to infinity of 1/n^2 converges.

When is each kind of statement used?

• Theorems are main results with broad applicability.
• A lemma is an intermediate result used to prove a theorem. An intermediate result may be stated and proved as a lemma rather than within the proof of the theorem if the result works for a more general context than that needed in the proof of the theorem within which it’s used, or if pulling the lemma out of the theorem proof makes the theorem’s proof easier for readers to understand.
• A claim is usually smaller than a lemma. It may appear within a proof to help structure the proof, or it may appear within exposition. It may be set formally using a claim environment or it may simply be signaled by wording such as “We claim that…”
• A proposition is a secondary or intermediate result, but is more interesting and valuable elsewhere than is a lemma. One frequently has a major result (a Theorem) whose proof is broken into many propositions, none of which is as interesting on its own as is the theorem.
• A corollary is a straightforward extension of a theorem; it is easily proved by using the theorem

Clearly a claim is something you must show; it is not self evident that the limit diverges. This is the justification. I know this is not quite related to "claims are not evidence" but I am thinking even more broadly now; claims are not support for anything. Once the claim is proven (in a mathematical setting) or substantiated with evidence (in an empirical setting), it assumes a new status of presumable fact. With regards to empirical reasoning, all claims have the evidential burden. Of course, there are many instances in your life where claims are all you have; it is not irrational to accept a proposition on the basis of your common sense (I mean this in the strict sense of the term) case-based reasoning, and useful heuristics. But let's not confuse the two concepts; a claim is not itself support for anything. I will repeat; a claim is a sentence that can be true or false, it is conceptually equivalent to a statement or proposition, they are bivalent, and there are different locutionary acts that may seem like claims but do not express a proposition (Like a question, command, statement of preference).

One more thing that really just bothers me: Absurd uses of Bayes theorem.

Bayes theorem represents the proper way to update posterior beliefs when confronted with data or evidence. It is simply a transformation rule that forces you to remain in accordance with the probability axioms. It says nothing about whether a particular piece of data is relevant, or what counts as evidence in the first place. This is an ontological distinction that is consistently ignored. If someone says “claims are evidence” and justify this categorization by reference to Bayes theorem, they are grossly misunderstanding Bayes theorem, committing a grotesque category mistake between the ontological status of evidence versus the process by which an agent rationally updates their beliefs once the evidence is established as such.  Take the hypothesis “the economy is in a recession”. There is evidence that confirms or disconfirms this claim; macroeconomic data related to the current state of the system (measurements economists deem relevant to diagnosing the health of the economic system). The hypothesis follows from the data whether we will be in a recession (or are in a recession). Now suppose we consult twitter and see some person claiming we are in fact in a recession. You search twitter and find 100 more independent accounts claiming we are in a recession. Does this count as evidence in favor of the hypothesis? No! This tells you something about consumer sentiment, and it’s true that sentiment has a bit of a self-fulfilling prophecy effect, but these are just claims. No matter how many claims you have about people’s perception of recession, the state of the system is confirmed by measurements with respect to the system. Plugging in to Bayes Theorem P(H=Recession|E=100 Claims on Twitter) is simply a non-starter; the data is irrelevant; and more broadly you cannot establish the health of any system by means of claims, only measurement. If you think you claims are evidence by reference to Bayes Theorem, you are making a category mistake. If we take the loosest definition of evidence, it is still the case that a claim itself is statistically independent of the hypothesis; it is the data contained in the claim.