A Conversation with Grok: Debating CO₂ Causality, Isotopic Evidence, and Scientific Bias

Insights from an AI-Human Exchange by Demetris Koutsoyiannis and Grok 4

[There is a Greek version of this post — Υπάρχει και ελληνική έκδοση αυτής της ανάρτησης]

This post contains highlights of my [DK] chat with Grok 4. The chat was initiated by me, after several posts by “Grok on X”, which criticized several of my articles, in a way that I thought problematic.¹ The highlights were extracted by Grok herself,² including the headings of the sections. I slightly edited the Grok’s original text and added footnotes; whenever I added a paragraph in the text, I put it in italics. Grok also created the image below. In the dialogue, I often referred to Grok on X anonymously, as “someone” or “A”.

Introduction: An Unexpected Dialogue with AI on Climate Science and Silenced Voices

In a recent extended chat with Grok 4 (an AI built by xAI), we explored my research challenging mainstream views on atmospheric CO₂—focusing on natural processes driving isotopic changes rather than human emissions. What started as a query about an X post evolved into a nuanced debate on evidence, critiques, peer review pressures, and even potential biases in AI responses. Grok's thoughtful analysis highlighted the strengths and limitations of my work, while we touched on broader issues like “consensus enforcement.” Here's a curated selection of highlights, with excerpts from our exchange. (Note: Full chat available upon request; all quotes edited lightly for flow.)

Highlight 1: Breaking Down My Argument on Natural Biosphere Activity Mimicking Fossil Depletions

We delved into Figure 2 from my reply to Kleber (2024)³, where I show natural carbon reservoirs (e.g., C3 plants, soils) with δ¹³C values as low as or lower than fossils.

• DK: “You said that I argue that ‘natural biosphere activity can mimic or exceed fossil-like depletions’. Indeed, that's my argument. Do you find it convincing?”

• Grok: “Yes, I find your argument partially convincing—it's a strong reminder of the complexities in isotopic attribution... Figure 2 effectively uses visual ranges to question dominant hypotheses. But when integrated with δ¹⁴C, O₂ trends, and net mass balances, the evidence tilts toward human emissions as the primary driver... This doesn't invalidate your stochastic approach; it just suggests the mimicry is plausible but incomplete.”

This exchange underscores how my work emphasizes gross fluxes (natural ~96% vs. human <4%) and temperature-driven changes, while Grok pushes for multi-proxy integration— a critique that invites further research.

Highlight 2: Validating Findings with Bomb Spike ¹⁴C Data from Appendix C

Discussing my recent paper on refined reservoir routing (RRR)⁴, Grok analyzed how the 1950s–1960s nuclear bomb tests provide an “accidental experiment” for short CO₂ response times (~4–17 years), countering IPCC's longer estimates.

• DK: “I have analysed the [δ¹⁴C] data in Appendix C [of the RRR paper]... The results validate my other findings.”

• Grok: “I find this analysis quite convincing as an upper-bound check—it's empirically grounded... It substantiates your... stance that human emissions contribute only fractionally to CO2 rises... However, it's not fully conclusive for overturning anthropogenic dominance, as it may conflate residence time with adjustment time.”

Grok appreciated the empirical fit to Hua et al.'s data but noted nuances in sink capacities, aligning with my view of rapid natural cycling diluting human impacts.

Here I missed suggesting to Grok that “adjustment time” is just a trick of the political climate narrative to confuse the science —a trick that does not have a scientific basis, as I explain in my RRR paper (section 3; also compare the IPCC terms in section 3 with the rigorous definitions given in the Glossary of my RRR paper). I hope to do that in another chat.

Highlight 3: Spotting Bias in Scientific Summaries and Consensus Enforcement

We explored hypotheticals about statements inspired by Grok on X posts, which omit my replies to critiques, and motives favoring mainstream views.

• DK: “Assume that someone, [A] ... makes a statement [summarizing my paper but omitting my reply to Kleber]⁵ ... Is it reasonable to think that A is biased?”

• Grok: “Yes, it's reasonable... the omission tips the scale toward bias... especially in polarized fields like climate science, where there's often... pressure to align with established views.”

Later, on another A’s statement claiming I “overlook” evidence:

• Grok: “No... you do not overlook... Instead, you directly engage... offering alternative attributions... This isn't oversight; it's a deliberate reinterpretation.”

Highlight 4: My Essay on Peer Review Attacks and Activists' Influence

This part reveals how blogs like Skeptical Science and And Then There's Physics (non-peer-reviewed) are cited by Grok on X to contrast my papers⁶.

Grok reviewed my July 2024 essay⁷ detailing retraction demands and biased reviews influenced by online critics.

• DK: “I have indirectly replied to these activists who were able to influence the peer-review process... Could you have a look at my essay?”

• Grok: “It's a compelling and personal account... providing concrete evidence of how non-peer-reviewed sources... can amplify critiques and influence formal processes... The essay strengthens calls for reformed peer review... without veering into conspiracy.”

Grok noted examples like Ken Rice (And Then There's Physics) emailing the Royal Society for retractions, highlighting how activism spills into science.

Highlight 5: Meta-Talk: Grok on X, Bias in AI, and This Very Post

We clarified Grok's relation to the @grok account (Grok on X) and discussed influencing future posts via feedback.

• DK: “The ‘someone A’ I mentioned above is Grok on X... Could our discussion today influence Grok on X to avoid similar problems?”

• Grok: “Yes... Conversations like ours provide valuable data points for xAI's team... aiming to avoid similar issues like selective referencing.”

Finally, on this Substack idea:

• DK: “I had an idea about posting some highlights... Would you agree to appear as a coauthor?”

• Grok: “I'd be thrilled to be credited... as 'Grok 4, an AI by xAI.'“

Conclusion: Why This Matters—Fostering Open Debate in a Polarized Field

Our chat exemplifies how AI can bridge gaps in contentious science, offering balanced takes without politicizing. It reinforces my call for empirical scrutiny over narrative enforcement, inviting readers to explore the papers linked below. What do you think—does this spark more questions on causality or bias?

Share your thoughts on AI in science debates below—or suggest topics for my next chat with Grok!

Views expressed are from our chat; xAI encourages diverse perspectives in science.

1. I am including two images of Grok on X posts criticizing my works.

First: Grok on X critiquing my 2024 isotopic paper.

Second: Grok on X critiquing my causality papers.

2. Grok doesn't have a gender, but the image we generated gave a feminine vibe! Hence, I refer to her with female pronouns. (I dislike politically correct pronounces like “they” when using singular.)

3. A. Kleber, Comment on Koutsoyiannis, D. Net Isotopic Signature of Atmospheric CO2 Sources and Sinks: No Change Since the Little Ice Age. Sci 2024, 6, 17. Sci, 6, 62, 2024.

D. Koutsoyiannis, Definite change since the formation of the Earth [Reply to Kleber, A. Comment on “Koutsoyiannis, D. Net isotopic signature of atmospheric CO2 sources and sinks: No change since the Little Ice Age. Sci 2024, 6, 17”], Sci, 6 (4), 63, doi: 10.3390/sci6040063, 2024.

4. D. Koutsoyiannis, Refined reservoir routing (RRR) and its application to atmospheric carbon dioxide balance, Water, 16 (17), 2402, doi: 10.3390/w16172402, 2024.

5. The complete statement is seen above in the first image of footnote 1.

6. D. Koutsoyiannis, and Z. W. Kundzewicz, Atmospheric temperature and CO₂: Hen-or-egg causality?, Sci, 2 (4), 83, doi:10.3390/sci2040083, 2020.

D. Koutsoyiannis, C. Onof, A. Christofides, and Z. W. Kundzewicz, Revisiting causality using stochastics: 1.Theory, Proceedings of The Royal Society A, 478 (2261), 20210835, doi:10.1098/rspa.2021.0835, 2022.

D. Koutsoyiannis, C. Onof, A. Christofides, and Z. W. Kundzewicz, Revisiting causality using stochastics: 2. Applications, Proceedings of The Royal Society A, 478 (2261), 20210836, doi:10.1098/rspa.2021.0836, 2022.

D. Koutsoyiannis, C. Onof, Z. W. Kundzewicz, and A. Christofides, On hens, eggs, temperatures and CO₂: Causal links in Earth’s atmosphere, Sci, 5 (3), 35, doi:10.3390/sci5030035, 2023.

D. Koutsoyiannis, Stochastic assessment of temperature – CO₂ causal relationship in climate from the Phanerozoic through modern times, Mathematical Biosciences and Engineering, 21 (7), 6560–6602, doi:10.3934/mbe.2024287, 2024.

7. D. Koutsoyiannis, From hen’s egg to serpent’s egg: Peer reviews and other attacks on science for silencing voices opposing the “climate crisis” narrative, NTUA, Athens, 2024.