Questions & answersPurpose, scores, evidence and corrections
- What is Slugfester for?
Slugfester helps readers examine the reasoning in public debates. Scorecards bring arguments, replies, critiques and source passages together so you can inspect an exchange, compare assessments and decide where a case needs more support. The project also explores how well artificial intelligence (AI) can perform this work. Its value rests on the usefulness and accuracy of the analysis.
- Why use AI?
AI makes detailed, structured analysis across many debates practical. It can help identify overlooked replies, compare reasoning and suggest stronger arguments for either side. Readers can assess those contributions on their merits without endorsing AI in general. AI authorship alone neither establishes nor refutes a conclusion; Slugfester remains responsible for what it publishes.
- What does a score mean?
A score estimates the quality of the reasoning presented under the assessment criteria. It considers support, logical coherence, responsiveness and related features explained in the assessment guide. An 80/100 score is not an 80% probability that the speaker is right. Seeking truth is the aim; scores help direct attention to the arguments and evidence. They do not settle a worldview or measure a person’s worth.
- What evidence is there that the assessments are accurate?
The linked recordings let readers check whether quotations, summaries and critiques fit the actual exchange. Another encouraging sign is visible in an interlocutor’s score graph: when their scores cluster within a narrow range across several debates, the assessments repeatedly place their debating competence at a similar level.
Under the standard blind-review workflow, the reviewers are independent AI agents called by the administrative AI. They do not receive earlier scorecards or published results, so those scores are not supplied as targets. Earlier assessments may use different controls.
This consistency supports confidence that the assessments are identifying a recurring pattern in the person’s performance. Accuracy still depends on how well the assessments represent and evaluate the actual arguments; a consistent bias or misunderstanding can also recur.
- What does a tight cluster of one person’s scores show?
It shows that the person was assessed as demonstrating a similar level of debating competence across those debates. On an interlocutor’s profile, the score graph groups their debate scores into five-point ranges. Taller bars mean more debates received scores in that range. When most scores sit in a few neighboring ranges, you can see the consistency at a glance.
For example, Tom Jump’s 6 published one-on-one scores are 74, 76, 78, 78, 80 and 80 out of 100. They span 6 points. That cluster indicates broadly similar assessed performance across these debates.
Tom Jump · 6 debatesOverall score / 100 · bar height counts debates To examine another person, select their name on the rankings page and look at their score distribution and debate list. Read the cluster alongside the number of debates, topics and opponents. It describes their assessed performance in those debates; it does not establish identical ability in every setting or guarantee that each judgment is correct.
- How much weight should I give rankings and small score differences?
Read an average alongside the number of debates, their topics, the opponents and the assessment procedures. The catalogue is curated, and speakers face different conditions. Repeated differences across several debates are more informative than one result; neighboring averages can be too close to support a confident ordering. The ranking study explains both the stable patterns and the uncertainty about exact places. A one-point lead by itself does not establish a reliable advantage.
- Can an assessment misunderstand an argument or reproduce bias?
Yes. A transcript can omit a qualification, a speaker can be misidentified, or an assessment can miss a reply or misread a disputed term. Separate AI reviews and fixed criteria help check these problems, but review agents using the same model can share a mistake. A detailed explanation makes a judgment easier to examine; it does not guarantee that judgment is correct.
- Did the participants say everything shown on their scorecards?
Distinguish the quoted source passages, the site’s summaries and critiques, and the explicitly labeled AI contributions. A steelman is a strengthened version of an argument; it can add reasoning the participant did not supply. These proposed improvements and new ideas are separate from the recorded performance and do not contribute to participant scores. An accurate account of the original argument is still essential.
- Can I inspect the prompts and instructions behind an assessment?
The guide below explains the standard workflow, and scorecards provide source links and assessment notes. A model’s name alone does not specify the instructions or evidence it received. Exact prompts and complete run records are not currently presented as a download for each debate. Use Feedback to identify the assessment and the materials you want to inspect. The current guide should not be assumed to describe every historical assessment.
- What if an assessment misrepresents me or contains another error?
Use the scorecard’s “Report a possible scorecard issue” link, or the Feedback form. Identify the page and disputed wording, explain the problem, and include the relevant source passage or timestamp where possible. You do not need to diagnose the AI or propose a replacement score. A specific objection can be valid regardless of whether its author likes the result. Errors may not be corrected immediately, and submission does not guarantee a reply or a change to the current scorecard.
- How can I contribute to the improvement of the scoring method?
Suggest an improvement to the choice of arguments, scoring criteria, weights or assumptions, and explain how it would improve the assessment. A methodological objection can matter across many debates, even when no quotation is wrong. You can also explore alternative scoring weights using the saved assessments.
Errors identified may not be corrected immediately. All feedback will be considered, but only complaints judged valid will be incorporated into the prompts used for future reevaluations of the entire debate corpus. Valid complaints will be generalized into review instructions and safeguards that apply to every assessment and score. The reviewers are independent AI agents called by the administrative AI. They will be asked to check for the underlying problem across all debates, so a valid complaint about one case can improve the process throughout the collection.
- Will the debates be reassessed as AI improves?
A site-wide reassessment is tentatively planned for spring 2027, subject to stronger AI agents being available. A newer model still needs to demonstrate better performance. Future improvements do not establish the accuracy of today’s assessments, and specific errors can be reported now.
From debate to scorecard
The standard one-on-one workflow · September 2026. Earlier scorecards may use different controls. Select a panel to enlarge it.
The administrative AI coordinates the process and calls independent AI agents to review the evidence and resolve disagreements. Throughout this guide and its diagrams, reviewers and adjudicators are those AI agents.
Prefer text? Read the complete guide
Controlled inputs, not perfect objectivity. Reviews by independent artificial-intelligence (AI) agents called by the administrative AI reduce avoidable influence, but shared model biases and mistaken judgments can remain. Selection is curated, not a representative sample of all debates.
Define the debate
Check for duplicates and identify the recording, speakers, event year, and a neutral central question. Record what each side must establish: its burden. A critic need not prove an alternative unless that burden is adopted. A duplicate recording is not silently reassessed.
Secure the complete source
Check transcript coverage, material gaps, and speaker changes against the full substantive debate. Use complete captions; transcribe if needed. A complete automatic caption track may be preferable to an incomplete human-written one. Preserve timestamps, acquisition information, and digital fingerprints that reveal later changes. Freeze the evidence, assessment rules, model settings, permitted repairs, and cost limits before judgment.
Map the arguments, then lock the evidence
Identify important claims, replies, concessions, and unanswered objections. Tie each selected move—a substantive argumentative contribution—to an exact source passage, speaker, topic, and relevant burden or reply. Section weights total 100%; move importance is 1, 2, or 3. Freeze both before scoring. Explain unequal coverage; never invent symmetry or drop arguments merely to fit the display.
Two isolated AI reviews
The administrative AI calls two independent AI agents. Both review the same locked evidence and rate every selected move, with reasons. Neither sees the other’s judgments, previous scorecards, published commentary, winner labels, or AI-added arguments. “Independent” means separate AI contexts with controlled inputs, not necessarily different models or model providers.
Resolve differences
Check audio whenever speaker attribution is below high confidence. Fixed rules flag dimension differences greater than 8 points, move-score differences greater than 4 points, adjustment differences greater than 2 points, and relevant categorical differences. The administrative AI calls a fresh independent AI agent to resolve disagreements—the adjudicator. This agent reviews the disputed evidence and anonymous options, without the review agents’ identities or calculated overall totals. Use the selected option for disputed ratings and rounded means for undisputed numerical ratings. Preserve both original reviews, the resolutions, and audio evidence, then freeze the resolved record.
Calculate the scores once
Six fixed dimensions determine each move’s score:
- 25% logical coherence: Does the conclusion follow from the premises?
- 20% evidence and warrant: Are the evidence and its connection to the conclusion adequate?
- 20% responsiveness: Does the move answer the actual live challenge?
- 15% relevance and burden: Does it advance or challenge what must be established?
- 10% precision and clarity: Are terms, scope, and qualifications clear?
- 10% calibration and charity: Is confidence proportionate, and is the opposing argument represented fairly?
Move: round the weighted sum of the six ratings. Section, for each side: multiply each move score by its previously locked importance, add these results, divide by the total importance, and round. Overall, for each side: combine section scores using their prelocked percentage weights, add the burden-completion adjustment, then round and keep within 0–100.
The default final adjustment is zero. An integer from −5 to +5 is allowed only for consequential full-debate burden completion not already counted in the moves. It cannot reward style or penalize the same defect twice. Models supply judgments; software calculates totals. No person hand-adjusts the result.
Two separate worked examples—not actual debate results: moves scoring 80 and 70 with importance 2 and 1 produce (80 × 2 + 70 × 1) ÷ 3 = 76.67, rounded to 77. Sections scoring 80 and 70 with weights 60% and 40%, plus a +2 adjustment, produce 48 + 28 + 2 = 78. The first example’s 77 is not an input to the second.
Stability checks: compare the final result with both original passes. Mean absolute distance is limited to 4 points, maximum distance from either pass to 8, and excursion beyond their range to 3. If both initial integer-rounded results favor the same side, the final result may favor that side or tie, but cannot reverse the winner. A failed check blocks acceptance; it is not permission to tune scores.
Explain the locked result
Publish every selected move exactly once. Quotations must match the eligible source wording and timestamps must lead to the relevant passage. Each critique explains the strongest feature, principal limitation, remaining burden, and locked score. Overall commentary identifies concrete strengths and material blunders for both sides. Explanations cannot change the selected moves, resolved judgments, or calculated totals.
Review fallacies and cognitive biases separately
After scores and critiques are locked, the administrative AI calls two fresh independent AI agents to examine every selected move against the local LogFall and CogBias definitions. Neither sees existing labels or the other agent’s review. The administrative AI then calls a fresh independent AI adjudicator to decide every candidate raised by either review agent, including candidates raised by only one.
A label must identify a material defect supported by the complete source and already reflected in the critique. A contested premise is not automatically circular reasoning; a weak analogy is not automatically equivocation; citing an expert is not automatically an appeal-to-authority fallacy. Distinct labels require distinct failures. When uncertain, omit the label.
Labels never add another numerical penalty. Finding no applicable labels is valid. Comparisons with earlier debates check consistency, not a target number of tags. Accepted labels link to local definitions and debate-specific explanations, with external LogFall or CogBias links for further reading.
Keep AI contributions separate
Clearly label strengthened arguments and genuinely new ideas for both sides as AI contributions. They are not transcript statements, are not attributed to the speakers, and never affect participant scores. Check their source-specific reasoning, added value, and repetitive wording. Biographical research also stays outside judgment packets and score calculations.
Verify before publication
Check source fingerprints, the score calculation chain, exact quotations, move coverage, label decisions, and the audit record. Compare explanation depth and specificity with 25 earlier debates without targeting their scores, argument counts, or tag rates. Test desktop and mobile pages, keyboard access, expandable sections, overflow, and links. Preserve rejected outputs and allow only bounded repairs that cannot rewrite locked scoring substance. Required checks must pass before publication, and the published result is then verified.
Team debates and scope limits
The ordinary workflow assesses one substantive advocate per side. A narrowly defined exception permits incidental host advocacy only when the recording is organized around two primary opponents, non-primary advocacy occupies at most 5% of the assessed window and adds no unique load-bearing argument, and excluded interventions and directly dependent replies are frozen and disclosed before judgment. Neither primary side receives credit or penalty for excluded material.
Panels and roundtables are not silently forced into that route. Authorized, stable two-team debates use a separate, disclosed approximation method. The team is the scoring unit: no individual score is inferred. Moves and labels retain their actual speaker attribution. Team and panel scores are excluded from individual rankings, profile averages, score distributions, and opponent records. Team checks include audio verification and tests of whether the leading side changes when contributions are rebalanced or one teammate is removed.
What scores mean: an estimate of argumentative performance in this debate, not the truth of a worldview, a speaker’s overall expertise, or a person’s worth. Fame, applause, accent, and speaking polish earn no automatic bonus. A true claim can be defended badly, and a false claim argued carefully. These safeguards improve consistency and traceability, not perfect objectivity.
Explore the assessments
The live distribution graph is available with JavaScript enabled. Read the research methods and limitations, browse all debate summaries, or report a possible scorecard issue.
Have a debate in mind? See which debates qualify and suggest a debate on the Feedback page.
Patterns across the debates
Charts
What gets argued. How it scores.
Where the reasoning is tested.
Explore arguments supporting and challenging theism and religious claims, using the saved SLUGFESTER assessments.
Showing the saved God & theism view at threshold 70. Filters, the threshold slider, and further evidence pages require JavaScript. Download all snapshot data.
Supporting and Challenging describe the positions argued on the selected religious question. They do not mean that every supporting move proves God, or that every challenge denies God.
Which arguments recur?
Share of debates containing at least one assessed move in each family, by position.
Morality & value
Meaning & society
Mind, reason & logic
Evil & suffering
Cosmology & contingency
Scripture & miracles
Fine-tuning & design
Experience & belief
Divine hiddenness
Other / general reasoning
Each debate counts once per family and position. Families overlap. Counts cover assessed material, not every spoken argument.
How do argument families score?
Median score and middle half of debate averages, on a 0–100 scale.
Morality & value
Meaning & society
Mind, reason & logic
Evil & suffering
Cosmology & contingency
Scripture & miracles
Fine-tuning & design
Experience & belief
Divine hiddenness
Other / general reasoning
The vertical mark is the median; the shaded span contains the middle 50% of debate averages. This is observed spread, not a confidence interval.
Making a case. Answering objections.
Average scores for constructive arguments and replies, with equal weight for each represented debate.
Morality & value
Meaning & society
Mind, reason & logic
Evil & suffering
Cosmology & contingency
Scripture & miracles
Fine-tuning & design
Experience & belief
Divine hiddenness
Other / general reasoning
Both bars use a 0–100 scale. The two roles can include different debates and arguments. A score difference is not evidence that a rebuttal caused a change.
Where does reasoning need more support?
Share of moves scoring below 70 on each recorded dimension, averaged equally across debates.
| Family / position | Logical coherenceDo the conclusions follow from the reasons given? This checks for contradictions, missing logical steps, and changes in meaning—not whether the conclusion agrees with the reader’s beliefs. | Evidence & supportAre claims backed by relevant facts, reliable sources, or good reasons for accepting the starting assumptions? Repeating a claim or naming an authority does not, by itself, supply that support. | ResponsivenessDoes the speaker address the other side’s actual argument or objection? A reply should engage the key point, rather than change the subject or attack an easier version of the argument. | Relevance & burdenDoes this argument help establish the claim the speaker needs to defend? Even a true statement may not advance that task. Asking the opponent to disprove a claim does not establish it. | Precision & clarityAre important terms, claims, and qualifications clear and used consistently? Readers should be able to tell exactly what is being claimed, including its scope and limits. | Calibration & charityDoes the speaker’s confidence match the available support, and is the opposing view represented fairly? This combines avoiding overstatement, acknowledging uncertainty, and describing the other side’s arguments accurately. |
|---|---|---|---|---|---|---|
| Morality & valueSupporting · 100 debates | 33.6% | 72.3% | 6.2% | 17.1% | 3.2% | 44.6% |
| Morality & valueChallenging · 95 debates | 10.8% | 32.1% | 7.6% | 12.2% | 1.4% | 17.9% |
| Meaning & societySupporting · 85 debates | 23.9% | 67.1% | 10.3% | 23.4% | 7.9% | 33.1% |
| Meaning & societyChallenging · 74 debates | 8.3% | 31.6% | 6.2% | 18.4% | 1.8% | 17.1% |
| Mind, reason & logicSupporting · 76 debates | 40.6% | 80.1% | 12.5% | 17.4% | 6.3% | 51.7% |
| Mind, reason & logicChallenging · 79 debates | 9.2% | 20% | 5.7% | 10% | 0.7% | 11.4% |
| Evil & sufferingSupporting · 72 debates | 27.7% | 73.8% | 6.1% | 14.3% | 8.4% | 35.9% |
| Evil & sufferingChallenging · 80 debates | 7.1% | 25.8% | 5.2% | 11.1% | 0.3% | 17.4% |
| Cosmology & contingencySupporting · 69 debates | 28.3% | 55.2% | 3.3% | 13.7% | 5.7% | 40.8% |
| Cosmology & contingencyChallenging · 62 debates | 10.6% | 24.4% | 3.9% | 5.8% | 3.6% | 14.1% |
| Scripture & miraclesSupporting · 72 debates | 18% | 66.6% | 13.1% | 23.9% | 2.8% | 38.3% |
| Scripture & miraclesChallenging · 64 debates | 5.3% | 35.1% | 3.6% | 12% | 1.2% | 23.6% |
| Fine-tuning & designSupporting · 57 debates | 22% | 66% | 7.8% | 6.5% | 4.5% | 44.7% |
| Fine-tuning & designChallenging · 62 debates | 12% | 33.8% | 5.9% | 7.8% | 5.3% | 18.8% |
| Experience & beliefSupporting · 45 debates | 21.5% | 75.2% | 15.4% | 34.4% | 2.2% | 34.1% |
| Experience & beliefChallenging · 47 debates | 3% | 20.5% | 7.4% | 11.7% | 1.5% | 13.7% |
| Divine hiddennessSupporting · 23 debates | 13% | 86.6% | 9.8% | 26.1% | 2.9% | 20.7% |
| Divine hiddennessChallenging · 26 debates | 5.8% | 19.2% | 9.6% | 15.4% | 0% | 6.5% |
| Other / general reasoningSupporting · 110 debates | 22.8% | 57.2% | 8.4% | 19.6% | 3.1% | 30.6% |
| Other / general reasoningChallenging · 118 debates | 8.6% | 24.3% | 4% | 13.9% | 2.3% | 15.2% |
All cells use a 0–100% scale. Zero means no recorded scores below 70; a dash means no observations. Higher percentages identify more below-threshold scores, not a cause of those scores.
Follow the evidence
The arguments behind the charts
3,239 unique assessed moves in this selection. Select an argument family above to narrow the list. Entries follow debate order.
Debate 01 · William Lane Craig
Equal conventional temporal intervals make a beginningless past actually infinite through successive instantiation, which Craig argues cannot culminate in the present.
Cosmic Beginning, Actual Infinity, and the Nature of Time · 25:45 · Cosmology & contingencyDebate 01 · Alex O'Connor
Big Bang evidence establishes an early hot, dense state, not necessarily an absolute first moment, leaving a finite cosmic beginning scientifically uncertain.
Cosmic Beginning, Actual Infinity, and the Nature of Time · 11:05 · Cosmology & contingencyDebate 01 · William Lane Craig
A closed universe with an infinite past should have reached thermodynamic equilibrium, so its observed low entropy supports a finite cosmic history.
Cosmic Beginning, Actual Infinity, and the Nature of Time · 35:43 · Cosmology & contingencyDebate 01 · Alex O'Connor
The traversal argument lacks an obviously non-arbitrary temporal unit, while presentism complicates treating nonexistent past events as an actually infinite collection.
Cosmic Beginning, Actual Infinity, and the Nature of Time · 19:23 · Cosmology & contingencyDebate 01 · William Lane Craig
Life-permitting constants appear extremely improbable under accident, making intentional determination by an intelligent creator a potentially better explanation.
Cosmic Fine-Tuning and Intelligent Selection · 94:39 · Fine-tuning & designDebate 01 · Alex O'Connor
If a creator intended material reality, background conditions making it extraordinarily difficult may count against the proposed intelligent-selection explanation.
Cosmic Fine-Tuning and Intelligent Selection · 95:53 · Fine-tuning & designDebate 01 · William Lane Craig
An evolutionary history involving animal suffering may be necessary for the feasible world in which the greatest number freely attain eternal salvation.
Animal Suffering, Theodicy, and Feasible Worlds · 68:53 · Evil & sufferingDebate 01 · Alex O'Connor
Human-centered defenses such as free will, soul-making, the fall, and higher-order goods do not themselves explain extensive evolutionary animal suffering.
Animal Suffering, Theodicy, and Feasible Worlds · 61:54 · Mind, reason & logic / Evil & sufferingDebate 01 · William Lane Craig
Many animals may feel first-order pain without reflective self-awareness, making their suffering less psychologically complex than comparable human suffering.
Animal Suffering, Theodicy, and Feasible Worlds · 79:24 · Evil & sufferingDebate 01 · Alex O'Connor
A creator could preserve complex animal behavior and ecological functions without sentience, retaining evolutionary outcomes while avoiding conscious animal pain.
Animal Suffering, Theodicy, and Feasible Worlds · 75:18 · Mind, reason & logicDebate 01 · William Lane Craig
Human ignorance of feasible worlds weakens the inference from no known sufficient divine reason for suffering to the absence of any such reason.
Divine Goodness, Moral Knowledge, and the Distribution of Evil · 71:48 · Evil & sufferingDebate 01 · Alex O'Connor
Distrusting moral intuition about apparently divine cruelty threatens moral arguments for God that rely on human awareness of objective moral truths.
Divine Goodness, Moral Knowledge, and the Distribution of Evil · 83:01 · Morality & valueDebate 01 · William Lane Craig
Strong independent evidence for a morally perfect God would support concluding that apparently gratuitous suffering serves sufficient reasons unknown to human observers.
Divine Goodness, Moral Knowledge, and the Distribution of Evil · 98:05 · Morality & value / Evil & sufferingDebate 01 · Alex O'Connor
The evidential difficulty concerns suffering's seemingly arbitrary severity across birthplace and species, not merely the fact that some suffering exists.
Divine Goodness, Moral Knowledge, and the Distribution of Evil · 105:20 · Evil & sufferingDebate 01 · William Lane Craig
The resurrection hypothesis best explains four widely accepted facts about Jesus, supporting the identification of the creator with specifically Christian theism.
Christian Identification, Experience, and Cumulative Warrant · 6:34 · Scripture & miraclesDebate 01 · Alex O'Connor
Evil remains evidence specifically against an all-loving God without needing to refute every creator hypothesis or erase every independent theistic consideration.
Christian Identification, Experience, and Cumulative Warrant · 99:25 · Evil & sufferingDebate 01 · William Lane Craig
Personal experience of God can rationally ground Christian belief as properly basic, without requiring philosophical or historical arguments as indispensable foundations.
Christian Identification, Experience, and Cumulative Warrant · 9:14 · Experience & beliefDebate 01 · Alex O'Connor
Suffering bears directly on Christianity because its saving, personally concerned God differs importantly from an indifferent first cause that permits suffering without intervention.
Christian Identification, Experience, and Cumulative Warrant · 101:52 · Cosmology & contingency / Evil & sufferingDebate 02 · William Lane Craig
Craig contends that no successful argument has established atheism and reserves his response until Hitchens presents arguments against God's existence.
Burden of Proof and Religious Epistemology · 15:04 · Other / general reasoningDebate 02 · Christopher Hitchens
Hitchens denies any duty to disprove God absent convincing evidence requiring that hypothesis and says observable phenomena remain explicable without it.
Burden of Proof and Religious Epistemology · 41:41 · Other / general reasoningRead the chart correctly
What this snapshot measures
Positions, not personal identities
Supporting and challenging refer to the religious claim at issue in each debate. They are assigned from the published positions, not from the stored PRO/CON columns or a speaker’s personal beliefs. Challenging a claim need not assert that no God exists. A creator-only argument does not establish a personal, intervening God.
Descriptive argument families
Families use explicit word and phrase rules applied to the assessed move and its identifier, with the section title as a fallback. A move can appear in several families; percentages therefore need not add to 100. These are browsing labels, not a fresh philosophical assessment. “Other / general” retains unmatched material.
Each debate gets equal weight
Score charts first average the unique assessed moves within each debate, position, family, and selected role. The score-range chart shows the median and middle half of those debate averages, not confidence intervals. The dimension grid averages each debate’s share below the selected score threshold (70 by default). Repeated speakers can still influence the sample; these figures do not establish statistical significance.
Saved assessments, limited sample
The snapshot includes 226 of 300 published debates available when it was built. It describes selected assessed moves, not every utterance or all public debates. Unclear position pairs, team approximations, and legacy panels without compatible argument-level records are excluded. Small samples are marked below five debates. Scores assess the reasoning presented, not whether God exists.
Classifications and exclusions
No cases awaiting AI classification review. All 74 exclusions have recorded decisions.
Constructive and reply groups can contain different debates. Their comparison does not measure an argument’s change after rebuttal. Concessions remain in the other charts but are omitted from the case/reply comparison.
Position assignments use the corrected September 1 research taxonomy where available, plus explicit AI reviews of previously omitted panels and later debates. Legacy grouped scores without compatible argument-level records remain excluded. Scope follows the main debate topic, with explicit editorial corrections. Individual arguments can address related questions within that debate.
Calibration concerns whether the strength of a claim matches its support; charity concerns fair representation of an opponent. Relevance and burden concerns contact with the question a side needs to establish.
- Debate 04: religion-adjacent, but no explicit theist-vs-non-theist side comparison
- Debate 18: The supporting column combines Williams, Goff, and Oldfield, including different concepts of God. These legacy panel scores lack the individual argument records and dimension scores required by the charts.
- Debate 19: no position-level theist-vs-non-theist comparison
- Debate 23: no position-level theist-vs-non-theist comparison
- Debate 32: Two-versus-two debate about the biblical God. The grouped Knechtle and O’Connor/Halper scores lack the individual argument records and dimension scores required by the charts.
- Debate 34: no position-level theist-vs-non-theist comparison
- Debate 35: The religious defense is in the CON column, opposite Carroll and Shermer. This team assessment lacks the individual argument records and dimension scores required by the charts.
- Debate 38: no position-level theist-vs-non-theist comparison
- Debate 40: no position-level theist-vs-non-theist comparison
- Debate 41: no position-level theist-vs-non-theist comparison
- Debate 44: no position-level theist-vs-non-theist comparison
- Debate 45: Moral realism combines secular and religious accounts on one side; it is not a pro-God position. This legacy panel assessment also lacks the required individual argument and dimension records.
- Debate 56: no position-level theist-vs-non-theist comparison
- Debate 71: Kushner and Gomes share a religious-humanist column opposite Hitchens. These grouped legacy scores lack the individual argument records and dimension scores required by the charts.
- Debate 73: no position-level theist-vs-non-theist comparison
- Debate 82: no position-level theist-vs-non-theist comparison
- Debate 84: Meyer and Lamoureux share the theistic CON column despite different accounts of design and evolution. These legacy panel scores lack the required individual argument and dimension records.
- Debate 95: The alternatives to Craig’s classical theism combine religious, God-ish, and skeptical approaches; they are not one non-theistic position. This legacy panel assessment also lacks the required individual argument and dimension records.
- Debate 96: Two-versus-two debate about Christianity, grouping the Knechtles against Aron Ra and Tom Jump. These legacy team scores lack the individual argument records and dimension scores required by the charts.
- Debate 105: Craig and Williams are grouped opposite Hossenfelder and Žižek. These legacy panel scores lack the individual argument records and dimension scores required by the charts.
- Debate 110: both published sides are theistic/religious
- Debate 111: no position-level theist-vs-non-theist comparison
- Debate 115: Horn and Bertuzzi are grouped opposite O’Connor and Schmid on the problem of evil. These legacy team scores lack the individual argument records and dimension scores required by the charts.
- Debate 125: Cliffe and Stuart Knechtle share the Christian-apologia column opposite O’Connor. These legacy grouped scores lack the individual argument records and dimension scores required by the charts.
- Debate 129: religion-adjacent, but no explicit theist-vs-non-theist side comparison
- Debate 131: no position-level theist-vs-non-theist comparison
- Debate 132: no position-level theist-vs-non-theist comparison
- Debate 133: no position-level theist-vs-non-theist comparison
- Debate 134: no position-level theist-vs-non-theist comparison
- Debate 135: no position-level theist-vs-non-theist comparison
- Debate 146: no position-level theist-vs-non-theist comparison
- Debate 149: Survival of consciousness after death does not itself assert God’s existence. This team assessment also lacks the individual argument records and dimension scores required by the charts.
- Debate 153: no position-level theist-vs-non-theist comparison
- Debate 154: The opposing column combines O’Connor’s secular critique and Kanojia’s spiritual inquiry; it is not uniformly atheistic. These legacy panel scores lack the individual argument and dimension records required for this comparison.
- Debate 159: religion-adjacent, but no explicit theist-vs-non-theist side comparison
- Debate 173: The motion concerns the Catholic Church’s institutional effects, with two speakers per side. These legacy team scores lack the individual argument records and dimension scores required by the charts.
- Debate 174: The published opposing column groups Dawkins and Kenny, rather than retaining separate participant records. This legacy panel assessment lacks the individual argument and dimension records required by the charts.
- Debate 180: both published sides are theistic/religious
- Debate 182: both published sides are theistic/religious
- Debate 183: no position-level theist-vs-non-theist comparison
- Debate 184: Moral realism versus anti-realism is not a theism-versus-skepticism comparison. This team assessment also lacks the individual argument records and dimension scores required by the charts.
- Debate 185: no position-level theist-vs-non-theist comparison
- Debate 186: no position-level theist-vs-non-theist comparison
- Debate 187: religion-adjacent, but no explicit theist-vs-non-theist side comparison
- Debate 188: no position-level theist-vs-non-theist comparison
- Debate 189: no position-level theist-vs-non-theist comparison
- Debate 190: no position-level theist-vs-non-theist comparison
- Debate 192: no position-level theist-vs-non-theist comparison
- Debate 193: both published sides are theistic/religious
- Debate 199: religion-adjacent, but no explicit theist-vs-non-theist side comparison
- Debate 201: religion-adjacent, but no explicit theist-vs-non-theist side comparison
- Debate 208: both published sides are theistic/religious
- Debate 213: no position-level theist-vs-non-theist comparison
- Debate 221: no position-level theist-vs-non-theist comparison
- Debate 224: no position-level theist-vs-non-theist comparison
- Debate 225: no position-level theist-vs-non-theist comparison
- Debate 226: both published sides are theistic/religious
- Debate 229: Historical existence of Jesus is not itself a theistic claim.
- Debate 230: Idealism versus naturalism does not establish a theism-versus-skepticism comparison.
- Debate 231: A soul claim is not itself a claim about God's existence.
- Debate 234: Two accounts of consciousness and purpose; no clear opposing religious and skeptical positions.
- Debate 235: A nonphysical soul is not itself a claim about God's existence.
- Debate 236: Moral realism versus anti-realism does not separate theism from skepticism.
- Debate 240: Alternative accounts of cosmic purpose and divinity.
- Debate 246: Alternative models of God; rejection of simplicity need not challenge theism.
- Debate 248: Dispute over classical-theist explanations rather than an unambiguous theism-versus-skepticism pair.
- Debate 249: Moral realism versus anti-realism.
- Debate 254: Team-debate approximation is excluded from this one-on-one comparison.
- Debate 257: Scope of science versus philosophy.
- Debate 259: Idealism versus illusionism.
- Debate 262: Alternative accounts of morality.
- Debate 264: Spiritual nondualism cannot safely be equated with theism.
- Debate 294: Both speakers argue within religious commitments; disagreement about conquest is not a theism-versus-skepticism pair.
- Debate 300: Compatibilism versus free-will denial.
Download the complete dated snapshot (JSON) · Research methods · Report a classification issue
These charts change only when a new snapshot is deliberately published. Adding or reassessing a debate does not refresh these charts. Source excerpts and scores here remain frozen; linked debate pages may receive later corrections.