• Tartas1995@discuss.tchncs.de
    link
    fedilink
    English
    arrow-up
    1
    ·
    3 days ago

    So you understand the difference between a claim and a question.

    And “is it true…?” Is a question, and while it might propose a claim, it doesn’t make it. The user doesn’t try to convince the llm of the claim. The user is challenging the reasoning of the answer of the llm. Consequently, the user doesn’t have a burden of proof. Llm has. There is no burden of proof move, because the user had it to begin with.

    • Zozano@aussie.zone
      link
      fedilink
      English
      arrow-up
      1
      ·
      3 days ago

      “is it true…?” Is a question that might propose a claim but doesn’t make it.

      Correct.

      The user doesn’t try to convince the LLM of the claim

      In the first question, no, but the followup conspiracy questions are poor attempts to do so.

      the user is challenging the reasoning of the LLM’s answer

      I’m not. You are. You still haven’t told me what it apparently got so wrong. It was supposed to be something horrible like ‘it will claim it’s impossible because that’s illegal’. For what it’s worth, I’ll chalk this exact claim up to hyperbole/exaggeration, but I don’t think it’s correct, at least on the Thinking models.

      the user doesn’t have a burden of proof.

      Not initially. The requirement for the need occurs when the contention arises.

      Burden of proof is only considered a logical fallacy because it diverts attention away from the argument and gives extra work to the listener.

      In this scenario we’ve discussed, the burden of proof arises once I notice that it’s raining.

      the LLM has the burden of proof.

      No it doesn’t, it’s not making a claim. It’s simply rejecting the claims I’m making because I cannot give it the evidence it needs to assess the validity of the proposition.

      there is no burden of proof move, because the user had it to begin with.

      Huh?

      the user doesn’t have a burden of proof.

      the LLM has the burden of proof.

      the user had it to begin with.

      • Tartas1995@discuss.tchncs.de
        link
        fedilink
        English
        arrow-up
        1
        ·
        edit-2
        3 days ago

        The follow-up conspiracy questions…

        Mhm… questions…

        Why does the llm assume that the user is making claims when the user challenges the llm’s reasoning?

        The user doesn’t need to make claims to challenge the llm’s reasoning. If the user asks questions without making claims, the user doesn’t have the burden of proof. At the time of the response of llm, the user hasn’t even challenge anything because the llm hasn’t answered the question, so there was nothing to challenge.

        So the user made no claim, and at the time, llm hasn’t made a claim. But when the llm answered the initial question, it made a claim. It got the burden of proof and the listener is the user.

        • Zozano@aussie.zone
          link
          fedilink
          English
          arrow-up
          1
          ·
          edit-2
          3 days ago

          Youre not picking up what I’m laying down.

          It is impossible to ask a question without making an implicit claim and asking another to verify.

          The reason the LLM is assuming the user is making claims is because that’s what a question is.

          This is a linguistic failure of English, as other languages drop the implicit claim entirely and just make it explicit.

          Surely you’ve heard non-native English speakers ask questions like “it is raining, is it not?”

          Sometimes English speaking people ask rhetorical questions like “isn’t the weather beautiful?”

          the LLM made a claim.

          What claim?

          the user made no claim.

          I/you/we did. We’ve been over this.

          • Tartas1995@discuss.tchncs.de
            link
            fedilink
            English
            arrow-up
            1
            ·
            2 days ago

            Your definition of claim is interesting as it is ridiculous.

            Let’s say for the sake of argument. That the question makes an implicit claim. Let’s say there is a claim.

            When I ask “does god exist?” And my answer would be “i don’t know.”, you would say that i made the implicit claim “god exists”, right? I will accept that i did for the sake of this argument.

            Is that claim one that I would have to prove? Do i have the burden of proof for the claim “god exists”? Am I moving my burden of proof onto someone else, if that person said “no.” And I would ask “how do you know?”?

            • Zozano@aussie.zone
              link
              fedilink
              English
              arrow-up
              1
              ·
              edit-2
              2 days ago

              A claim is not an assertion of fact.

              The burden of proof requirement is only required once there is contention. “I don’t know” doesn’t trigger it unless one person makes a positive proposition and the other person rejects it.

              • Tartas1995@discuss.tchncs.de
                link
                fedilink
                English
                arrow-up
                1
                ·
                edit-2
                2 days ago

                So in your opinion, the disagreement makes the claim require proof and consequently the person making the claim, has the burden of proof?

                Edit: sorry, I said disagreement but for clarity, does it have to be a disagreement? Or is the question, e.g. “how do you know?” Enough for proof to be required?

                • Zozano@aussie.zone
                  link
                  fedilink
                  English
                  arrow-up
                  1
                  ·
                  edit-2
                  2 days ago

                  So in your opinion, the disagreement makes the claim require proof and consequently the person making the claim, has the burden of proof?

                  Bingo

                  does it have to be a disagreement? Or is the question, e.g. “how do you know?” Enough for proof to be required?

                  The exact requirement is that the other person does not accept the claim. Being sceptical is enough - it doesn’t need to be a flat-out rejection.

                  Example: I tell you I have a pet snake, you say “do you really?”. If I want to convince you I have a snake, I need to provide evidence.

                  Example 2: I ask “does god exist?” And you reply with:

                  • I don’t know
                  • I don’t believe in god
                  • I’m not convinced there is a god
                  • maybe

                  These are fine. None of these are claims.

                  However, if you were to assert “god exists” or “god does not exist” then you’ve got an issue, and a burden of proof to meet.

                  Note: negative claims can require a burden of proof (but its not a good place to be in). Take my case:

                  I’m an atheist. I am not convinced God exists. It is a fallacy for me to claim as a fact “God does not exist” because that is a burden of proof I cannot meet without constraints (Precise definition, time, place). For example: Zeus does not exist in my bedroom right now.

                  Strangely enough, people who assert as a fact the claim “God exists” are in a slightly better position (but still not a good), because they can assert their definition of God is imperceptible and everywhere, or they met god on a train last weekend while zonked out of their gourd.

                  It’s my choice to believe their claim without contention, or ask them to prove their claim.

                  • Tartas1995@discuss.tchncs.de
                    link
                    fedilink
                    English
                    arrow-up
                    1
                    ·
                    edit-2
                    2 days ago

                    Sorry, misread something.

                    So you think that the person answering the question “does god exist?” with “no”, has the burden of proof if i would ask how?