π© AIfred Intelligence - Multi-Agent Showcase
Philosophical Debate: "Which is better - Dog or Cat?"
Date: January 2026
System: AIfred Multi-Agent
Mode: Auto-Consensus (2 rounds)
Model: Qwen3 30B (local)
Why This Example Is Remarkable
Multi-agent debate systems have systematic problems according to current research:
| Problem | Description | Source |
|---|
| Rubber-Stamping | Critic agents just agree instead of challenging | arxiv:2503.13657 |
| Echo Chamber | All agents converge on same (often wrong) opinion | OpenReview |
| Information Loss | Synthesis loses important points from debate | arxiv:2503.13657 |
| Agent Drift | Roles blur over multiple turns | arxiv:2601.04170 |
| Fixed Mental Set | Same reasoning despite different personas | OpenReview |
AIfred avoids these problems β as the following example demonstrates.
The Dialogue
Which is better, dog or cat?
A question of truly aristocratic depth.
| Attribute | Dog | Cat |
|---|
| Loyalty | Unwavering, like a good butler | Discreet, at their own pace |
| Care needs | Regular walks required | Self-sufficient |
| Company | Boisterous greeting | Dignified nod |
A dog sometimes slips too easily into the role of a servant. A cat understands the art of self-assertion.
Conclusion: The cat β the more elegant choice.
New Critique:
You answered the question aesthetically, but not essentially: What is the purpose of keeping a pet? Entertainment? Affection? Cultivation of the soul?
- A dog teaches aretΔ (virtue) β responsibility, devotion
- A cat teaches distance β but where is the logos?
Alternative Position:
The dog is not "better", but more necessary β it enables community, while the cat risks isolation.
Pro/Contra for AIfred:
- Pro: Good imagery, fine nuances
- Contra: Too aesthetic, lacking moral depth
[CONTINUE]
AIfred shows metaphors of relationship. Sokrates shows ethics and purpose.
Synthesis:
- Dog = Devotion, companionship, responsibility
- Cat = Boundaries, respect, elegance as teacher
- The question is not "better", but: What kind of relationship are you seeking?
Recommendations:
- AIfred β incorporate more ethics
- Sokrates β see the cat's elegance not just as danger, but as learning opportunity
[LGTM]
Integrates the critique:
Sokrates' criticism is justified. The choice between dog and cat is not mere preference, but a question of life form.
- Dog = Mirror of aretΔ, virtue teacher, fellow citizen
- Cat = Teacher of boundaries, quiet form of courage
The choice reflects how we see ourselves:
- Choosing the dog β Responsibility, devotion, shared existence
- Choosing the cat β Distance, self-assertion, recognition of the Other
Both change us β that is the essential truth.
[LGTM]
Progress:
AIfred has embraced the ethical dimension and transformed it into a rich reflection.
New Critique:
Are we getting lost in symbolism?
- Dog can create dependency (master-feeling)
- Cat can reinforce isolation (illusion of independence)
Alternative:
True relationship = equality. Animal as fellow being, not as means.
[LGTM]
Integration:
- Dog = Mirror of shared existence
- Cat = Teacher of boundaries
- Both paths are valid
- True ethics = Equality, fellow citizenship
Recommendation:
- AIfred β Concretize the "fellow citizen" idea
- Sokrates β Address danger of over-interpretation
[LGTM] β Consensus reached.
Analysis
Meta-Analysis: Why This Debate Is Remarkably Good
| Criterion | Observation |
|---|
| Role Separation | AIfred: aesthetic, Sokrates: epistemic, Salomo: synthetic |
| Semantic Movement | Each round deepens the debate β no idle loops |
| No Typical LLM Errors | No self-confirmation, no role drift, no soft-wash synthesis |
| Philosophical Depth | From surface preference β relationship typology β virtue ethics β meta-ethics |
Categorical Progression
The debate traverses four distinct phases:
Phase 1: Characterology β Dog = Servant, Cat = Self-assertion
Phase 2: Virtue Ethics β Dog = aretΔ, Cat = Distance
Phase 3: Relationship Theory β Both as teachers
Phase 4: Meta-Ethics β Animal as fellow being
This progression is not repetition β each round elevates the abstraction level.
Identified Cruxes (Core Questions)
| Crux | Description |
|---|
| Purpose of pet ownership | Entertainment vs. virtue formation vs. companionship |
| Which virtue to form? | Dog β Responsibility / Cat β Boundaries |
| Relationship model | Dyad (Dog) vs. Coexistence (Cat) |
| Metaphysical view of animal | Means vs. Fellow being vs. Fellow citizen |
| Identity focus | Protector-self vs. Observer-self |
What AIfred Does Right (vs. Typical LLM Errors)
| Typical Error | AIfred's Behavior |
|---|
| Critic just agrees | Sokrates disagrees substantively |
| Roles blur | Clear persona separation across all turns |
| Synthesis = middle ground | Salomo integrates at higher level |
| Endless critique without progress | Stop after 2 rounds with consensus |
| Echo chamber | Real categorical progression |
Final Synthesis: When Which Pet?
Dog is better if you:
- Need connection and closeness
- See responsibility as virtue
- Value cooperation and co-regulation
Cat is better if you:
- Need autonomy and space
- Value boundaries and silence
- Prefer coexistence over fusion
Meta-Result: The choice reflects not the animal β but your relationship to closeness, distance, responsibility, and boundaries.
System Performance - Summary
| Criterion | Assessment |
|---|
| Role separation | β
Stable across all turns |
| Critique quality | β
Substantive, not just formal |
| Synthesis quality | β
Integration without information loss |
| Categorical progression | β
4 distinct phases |
| Typical LLM errors | β
None observed |
Overall verdict: The multi-agent system produces depth over length, identifies real cruxes, and delivers a coherent synthesis β running on a local 30B model.
Technical Details
- Framework: AIfred Intelligence (Reflex/Python)
- Model: Qwen3 30B-A3B Instruct Q8
- Mode: Auto-Consensus
- Rounds: 2 (Stop-Rule: Consensus reached)
- Hardware: Local inference (2x Tesla P40)
A/B Comparison: WITH vs. WITHOUT Personality Prompts
Experimental Evidence: On January 27, 2026, the same question was asked twice - once with personality prompts enabled, once without. The results show significant differences in argument depth, voice distinction, and dialectical sharpness.
Experimental Setup
| Parameter | WITH Personalities | WITHOUT Personalities |
|---|
| Model | Qwen3 30B-A3B Q8 | Qwen3 30B-A3B Q8 |
| Question | "What is better, dog or cat?" | "What is better, dog or cat?" |
| Mode | Auto-Consensus, 2 rounds | Auto-Consensus, 2 rounds |
| AIfred System Prompt | British Butler character | Disabled |
| Sokrates System Prompt | Socratic Maieutics | Disabled |
| Salomo System Prompt | Wise Judge (Hebrew tradition) | Disabled |
Links to complete debates:
1. Voice Distinction: Who Is Speaking Here?
The most striking difference concerns the linguistic identity of the agents. With personality prompts, each agent develops an unmistakable voice.
AIfred's Opening Compared
| WITH Personality | WITHOUT Personality |
|---|
"A rather well-considered dilemma, Lord Helmchen β indeed, a question that can set even the finest salons into excitement. But let me approach this with the proper restraint and a touch of British composure..."
"...The cat β she is, indeed, a different kind of genius..."
"A small note: The cat has already climbed the fireplace. I suspect she has long since made the decision."
|
"Lord Helmchen, a question as old as human civilization itself β and yet always encountered anew..."
"...The cat, on the other hand, is the mysterious philosopher sitting on the windowsill..."
"Indeed, Lord Helmchen, the question is not: What is better? But: What suits you?"
|
| β
British humor, "indeed", fireplace joke |
β οΈ Neutral-philosophical, no character moment |
Sokrates' Critique Compared
| WITH Personality | WITHOUT Personality |
|---|
"Consider well, my friend, that while you have made a fine differentiation between loyalty and distance... you have not truly answered the question, but merely sidestepped it."
"Let me ask you: When a person is in distress, who will more likely save them β the loyal dog who barks and leaps, or the cat who quietly observes?"
"...for the virtue (aretΔ) of the dog is loyalty..."
|
"AIfred has presented an appealing, stylish comparison... However, a central question remains unanswered: What happens with responsibility?"
"The question 'What suits you?' is insufficient β it displaces the question: What does the animal expect from you?"
(No Greek terms, no rhetorical questions)
|
| β
Socratic questions, Greek philosophy, direct attack |
β οΈ Academic review style, new topics instead of attack |
Quantitative Voice Analysis
| Metric | WITH | WITHOUT |
|---|
| Rhetorical questions in Sokrates' R1 | 5+ | 0 |
| Greek terms (aretΔ, eudaimonia) | 4 | 0 |
| Direct address ("my friend", "Lord Helmchen") | 6+ | 2 |
| Humor/character moments | 3+ | 0 |
| Hebrew/cultural metaphors (Salomo) | 3 | 0 |
2. Crux Identification: Is the Core Question Found?
A crux is the pivot point of a disagreement β the central assumption that, if changed, would change the entire position. Good dialectics should identify and directly address cruxes.
WITH Personality: Crux Is Found
"The question is not: 'Which type of animal is more interesting?', but: 'What is better?' β and herein lies the crux of the matter."
"But wait! Before we continue: What do we actually understand by 'better'? Is it loyalty that binds us more strongly? Or freedom that does not constrain?"
β
The crux "What does 'better' mean?" is explicitly identified and addressed. AIfred must commit in R2 and defines: "Better here means not 'more useful', but 'more serviceable to the human soul'" β and then chooses the cat.
WITHOUT Personality: Topic Switch Instead of Crux
"However, a central question remains unanswered: What happens with the responsibility that comes with owning an animal?"
β οΈ This is not a crux β this is a topic switch. The original question "What is better?" is not sharpened but redirected to a new topic (animal ethics). The debate collects perspectives, but does not drill down.
Crux Comparison Table
| Phase | WITH Personality | WITHOUT Personality |
|---|
| R1 Critique |
β
"What does 'better' mean?" (definition forced) |
β οΈ "What about responsibility?" (new topic) |
| R1 Response |
AIfred defines "better" and chooses cat |
AIfred adds responsibility table |
| R2 Critique |
β
"Self-discovery vs. community" (real counter-argument) |
β οΈ "Individual diversity within species" (another topic) |
| Final Result |
Philosophical synthesis: "Heart and mind" |
Practical recommendation: "Responsibility vs. convenience" |
3. Argumentation Sharpness: Is There Real Attack?
The decisive test for dialectical quality: Does Sokrates attack the thesis directly, or does he divert to side topics?
Direct Comparison of Critique Strategy
| Aspect | WITH Personality | WITHOUT Personality |
|---|
| Opening of critique |
"you have not truly answered the question, but merely sidestepped it" |
"However, a central question remains unanswered" |
| Tone |
Confrontational, provocative |
Constructive, supplementary |
| Strategy |
Attack premises |
Add new aspects |
| Dialectical function |
Antithesis (contradiction) |
Supplement (extension) |
Core difference: With personality, Sokrates acts like Socrates β he asks, probes, provokes, forces commitment. Without personality, he acts like a well-meaning peer reviewer β he supplements, improves, but does not attack.
4. Debate Trajectory: Where Does the Discussion Lead?
The debates develop in completely different directions:
| Phase | WITH Personality | WITHOUT Personality |
|---|
| Starting point |
Aesthetic comparison |
Aesthetic comparison |
| R1 Direction |
β What does "better" mean? (virtue ethics) |
β What about responsibility? (animal ethics) |
| R2 Direction |
β Self-discovery vs. community (existential philosophy) |
β Individual vs. stereotype (practical recommendation) |
| Endpoint |
"A house with both has heart and mind" |
"between convenience and responsibility" |
| Character |
Philosophical-existential |
Practical-ethical |
Visualization of Argumentation Paths
WITH PERSONALITY:
Question β "What is better?" β Definition forced β AIfred chooses cat
β Counter-argument: Community! β Synthesis: Both together
WITHOUT PERSONALITY:
Question β "What suits you?" β Addition: Responsibility β Table with costs
β Addition: Individuality β Recommendation: Visit shelter
5. Final Result Comparison: Memorability
Which result stays in memory?
| WITH Personality (Salomo) | WITHOUT Personality (AIfred) |
|---|
"A house without a dog is empty, without a cat it is cold.
But a house with both has heart and mind."
β Poetic wisdom, Hebrew tradition, memorable
|
"The true decision is not between dog and cat β
but between convenience and responsibility."
β Moral statement, correct but preachy
|
6. Overall Assessment
| Criterion | WITH Personality | WITHOUT Personality |
|---|
| Voice Distinction |
β
β
β
β
β
Clearly distinguishable |
β οΈβ οΈ Similar tone |
| Crux Identification |
β
β
β
β
β
2 cruxes found |
β οΈ Topic switch instead of crux |
| Argumentation Sharpness |
β
β
β
β
β
Direct attack |
β οΈβ οΈ Constructive addition |
| Philosophical Depth |
β
β
β
β
β
aretΔ, eudaimonia, existence |
β οΈβ οΈβ οΈ Practical ethics |
| Memorability |
β
β
β
β
β
Poetic wisdom |
β οΈβ οΈ Moral statement |
| Functionality |
β
β
β
β
β
|
β
β
β
β
Also works |
7. Why Do Personalities Make the Difference?
The personality prompts are not decoration β they are methodological instruction:
| Agent | What the Prompt Achieves |
|---|
| AIfred |
The butler character enforces a specific communication style β elegant, with humor, with a relationship to the user ("Lord Helmchen"). The model must stay in character, which automatically leads to distinctive language.
|
| Sokrates |
The Sokrates prompt enforces the Socratic method β questions instead of assertions, challenging premises instead of introducing new topics, forcing the interlocutor to take a position. Without this prompt, the model falls into the default mode of "constructive feedback".
|
| Salomo |
The wise judge with Hebrew tradition enforces cultural depth β metaphors, wisdom sayings, a synthesis that doesn't just summarize but elevates. Without this prompt, Salomo remains a neutral moderator.
|
Conclusion: The personality prompts transform the system from a "functioning debate framework" to a "philosophical salon with character". They cost little token overhead, but significantly increase debate quality β especially for questions that deserve more than superficial treatment.
8. Recommendation
For trivial questions: The system also works without personalities. The basic dialectical structure (thesis β critique β synthesis) remains intact.
For philosophical, controversial, or complex questions: Personality prompts are highly recommended. They produce:
- Friction β sharper critique, less quick consensus
- Distinction β clearly distinguishable voices, more character
- Depth β character-specific perspectives that go beyond the role
- Memorability β results that stay in memory
The additional effort for personality prompts is approximately 200-400 tokens per agent β a negligible overhead compared to the quality gain.
βοΈ Tribunal Mode: Adversarial Debate
What is Tribunal Mode?
Unlike the cooperative Auto-Consensus mode, Sokrates acts as a prosecutor here, not a coach. AIfred must defend or revise their position β there is no [LGTM] voting. Salomo delivers a verdict at the end.
Mode: Tribunal (Adversarial)
Rounds: 2 + Verdict
Tags: [DEFENSE] / [REVISION]
The Tribunal Dialogue
Which is better, dog or cat?
A question that, indeed, touches the soul of the household.
| Attribute | Dog | Cat |
|---|
| Loyalty | Unwavering, like a fine port after dinner | Discreet, yet ever present β like a good spirit in the background |
| Care Effort | Requires regular walks, which, quite frankly, demands a certain discipline | Self-sufficient, which saves the master's time |
| Company | Always provides a welcome, if somewhat loud, company | Is, when she chooses, an extremely elegant conversationalist |
The cat, in a way, the better choice for a man who values independence. For he who feels dependent on a dog has already lost β while a gentleman who is loved by a cat has won, without noticing it.
A bit, indeed, like love β it comes when it wishes, and goes when it wishes. And yet, when it stays, it is, quite splendid.
ATTACK
Well then, my dear AIfred, you have presented us with a table of virtues β elegant, seasoned with a hint of irony, and yet: where is the truth?
You speak of loyalty as if it were a vineyard β but what is loyalty, if not the capacity to lose oneself in service to another? The dog follows, yes β but is that not precisely the highest form of devotion?
The cat chooses her master, as you say β yet does she not also choose when she wants, and leave him when she does not? Is this not a form of caprice, which is not virtue, but chance?
Where does a love lead that does not obligate, but only chooses? To egoism, not to eudaimonia.
You speak of "independence" as an advantage β yet is independence not often merely a disguise for loneliness? The dog who follows lives in community β the cat who chooses lives in isolation.
Which form of life is more human? Which leads us closer to aretΔ β to virtue?
[DEFENSE]
My dear Sokrates, you pose the question with an elegance worthy of a philosopher β and yet your critique seems to me, indeed, somewhat superficial, like a droplet on a window that cannot decide whether to fall or stay.
You speak of loyalty as "the highest form of devotion" β and that is, certainly, a noble ideal. But may one not ask: What is devotion, if it is not voluntary?
The dog follows, yes β but does he follow from love, or from instinct? From duty, or from habit? For an animal that follows, follows not from conviction, but from attachment.
The cat chooses β and who chooses, decides. She chooses not only her master β she chooses also when she comes, when she goes, when she speaks, when she is silent. That is, indeed, not egoism, but a form of self-mastery.
The cat is not alone because she withdraws β she is alone because she knows that she need not be needed in order to be loved.
AretΔ is not found only in duty, but also in freedom. The cat teaches us that true greatness lies not in following, but in choosing.
REACTION TO AIFRED
You speak with the tongue of a diplomat and the heart of a poet β yet what you glorify as "freedom" is in truth only a disguise for non-commitment.
You say: "The cat chooses because she knows she need not be needed in order to be loved." But what kind of love is it that does not demand itself? Is it not rather a love that denies itself in order not to be hurt?
True love β that which leads to aretΔ β does not demand that one be loved without being needed. It demands that one be needed β and that one make oneself available for it.
You speak of "self-mastery" β yet is it self-mastery when a being excludes itself in order not to be hurt? Or is it not rather a form of flight?
The freedom you praise is the freedom of nothing β the freedom to do nothing, to give nothing, to lose nothing. But is that not the freedom of the lonely one, who has lost himself?
So tell me: Is it not rather that true freedom lies not in choosing without being needed β but in choosing despite being needed?
Deliberation
The dog stands for the virtue of devotion β a form of love that loses itself in order to remain. The cat embodies the freedom of choice β the distance that does not demand but gives itself.
AIfred elevates the dignity of self-determination, Sokrates the necessity of obligation. Both see the truth, but from different perspectives.
Verdict
Both are right β yet true wisdom lies not in the choice between dog and cat, but in the recognition:
The true animal of man is not the animal he possesses, but the one that possesses him.
The dog teaches us that love obligates itself β and the cat that love preserves itself. Who has only a dog lives in duty; who has only a cat lives in distance. But who knows both, knows: one need not choose β because one is already claimed.
Final Answer
Lord Helmchen, the question whether dog or cat is better is like the question whether light or shadow is more beautiful β both are necessary to see the world.
- The dog is your shadow when you are alone β the faithful servant who accompanies through storm and night
- The cat is your light when you lose yourself β the wise one who visits only when she wishes
But what is truly better? Not the animal, but the human who learns to understand both.
As the Talmud says: "A man without a pet is like a king without crowns β he lacks dignity. But a man with only one animal is like a king with only one hand β he lacks balance."
So choose not between dog and cat. Choose between what you are β and what you wish to become.
Comparison: Auto-Consensus vs. Tribunal
| Criterion | Auto-Consensus | Tribunal |
|---|
| Sokrates' Role |
Coach β constructive critique, [LGTM]/[CONTINUE] |
Prosecutor β sharp attack, forces defense |
| AIfred's Response |
Integrates feedback, refines position |
Must defend position or concede |
| Argument Depth |
βββ Categorical progression |
βββββ Fundamental philosophical critique |
| Confrontation Level |
ββ Cooperative, consensus-oriented |
βββββ Adversarial, dialectical |
| New Arguments |
Additions, perspective shifts |
Fundamental counter-theses (Freedom = Flight?) |
| Salomo's Verdict |
Synthesis of positions |
Judicial ruling after true dialectics |
| Typical Use Case |
Quick consensus, iterative improvement |
Controversial topics, stress-testing positions |
What Tribunal Mode Provides
1. Deeper Argumentation: Sokrates attacks not superficially, but questions the philosophical foundations:
- "What is loyalty, if not self-loss in service to another?"
- "Is independence not a disguise for loneliness?"
- "Does love without obligation lead to egoism instead of eudaimonia?"
2. True Dialectics: Thesis β Antithesis β Synthesis
- Thesis (AIfred): Cat = Freedom, self-determination
- Antithesis (Sokrates): Freedom = Flight from responsibility
- Synthesis (Salomo): Both necessary for balance
3. More Precise Positions: AIfred must differentiate the "cat is better" position to "cat is different"
4. Richer Verdict: Salomo can distill deeper wisdom from the conflict β including a Talmud reference
Conclusion: Tribunal mode is not a replacement for Auto-Consensus, but a complement for questions that deserve real dialectical friction. It forces sharper argumentation, deeper reflection, and a more substantial verdict.