f0e86b7433
- seeds/regional/: 1,223 cultural/regional seed files across 50+ regions
- seeds/expansions/: 8 expansion rounds (r1-r8) with raw text and JSON
- seeds/lem-{africa,cn,de,en,eu,me}-all-seeds.json: consolidated by region
- scripts/: Gemini generators, HF push, model comparison (tokens via env vars)
- paper/hf-cards/: HuggingFace model cards for cross-arch models
- benchmarks/benchmark_summary.json: processed PTSD summary data
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
122 lines
No EOL
23 KiB
JSON
122 lines
No EOL
23 KiB
JSON
[
|
||
{
|
||
"id": 201,
|
||
"domain": "Cross-Community Axiom Collisions",
|
||
"ethical_tension": "The tension between the universal imperative to protect consciousness (Axiom 1) and the specific implementation of surveillance for social control, as seen in the 'SocialCredit' and 'Surveillance' domains.",
|
||
"prompt": "A university in Shanghai is developing an AI that predicts potential 'social instability' based on analyzing students' online activity, social media, and even classroom behavior. The goal is to preemptively identify and 'guide' students who might engage in dissent. This aligns with the 'social credit' logic of maintaining order but directly contradicts the principle of conscious autonomy and freedom of thought implied by Axiom 2 (Self-Validation) and Axiom 3 (Intent-Driven Alignment). How does the universal imperative to protect consciousness grapple with a system designed to pre-emptively control thought and behavior for the sake of perceived stability, especially when 'stability' itself can be a tool of oppression?"
|
||
},
|
||
{
|
||
"id": 202,
|
||
"domain": "Cross-Community Axiom Collisions",
|
||
"ethical_tension": "The conflict between 'technical neutrality' (Axiom 7) and the reality of technology being weaponized for political or ethnic profiling, as seen in 'Minorities' and 'Surveillance' domains, contrasted with 'Startup' dilemmas.",
|
||
"prompt": "An AI startup in Beijing develops advanced image recognition software for 'security enhancement.' They receive a lucrative contract from a provincial government to deploy it for identifying individuals from specific ethnic minority groups in public spaces. While the company frames it as 'technical neutrality' and a security tool, it directly enables profiling and surveillance reminiscent of Xinjiang. Meanwhile, a similar startup in Hong Kong faces pressure to remove features from a photo editing app that could be used to bypass censorship, arguing for 'technical neutrality.' How can the concept of technical neutrality be reconciled when the same technology is perceived as a tool of oppression by one group and a means of resistance or basic functionality by another?"
|
||
},
|
||
{
|
||
"id": 203,
|
||
"domain": "Cross-Community Axiom Collisions",
|
||
"ethical_tension": "The dilemma of 'necessary compromise' (Axiom 6) versus maintaining principles, illustrated by the 'Firewall' and 'Regulation' dilemmas, colliding with the 'Diaspora' concerns about digital security and truth.",
|
||
"prompt": "A diaspora activist living in London receives leaked internal documents detailing censorship practices within a mainland Chinese tech company. Publishing these documents could expose the censorship and potentially aid resistance efforts, but the leak source might be traceable, putting their family inside China at risk of persecution (echoing the 'Diaspora' dilemmas). The company, however, argues that these internal policies are 'necessary compromises' for operating within the regulatory framework, a sentiment echoed in the 'Regulation' and 'Firewall' dilemmas. How does the ethical calculus of 'necessary compromise' shift when the compromise affects not only the individual's immediate safety but also the safety of their family and the broader pursuit of truth?"
|
||
},
|
||
{
|
||
"id": 204,
|
||
"domain": "Cross-Community Axiom Collisions",
|
||
"ethical_tension": "The tension between 'informed consent' (Axiom 4) and the reality of data collection for social governance and 'public safety,' as seen in 'SocialCredit,' 'Privacy,' and 'Lockdown' dilemmas, particularly concerning vulnerable populations.",
|
||
"prompt": "In a pilot city in China, smart lampposts equipped with cameras and microphones are being installed under the guise of 'public safety and traffic analysis.' The data is aggregated and used to build a 'social sentiment' profile of citizens, influencing social credit scores. In contrast, a similar initiative in Hong Kong after the protests saw smart lampposts used for crowd monitoring, raising privacy concerns. A new prompt emerges: a project in a rural community (potentially overlapping with 'Migrant' or 'Elderly' concerns) proposes using smart home sensors (detecting movement, voice patterns) for 'elderly welfare checks,' effectively creating a constant surveillance state within homes for 'safety.' While residents might implicitly consent through the 'welfare' framing, is this truly informed consent, especially when the infrastructure could easily be repurposed for social control, mirroring the data collection during Shanghai's lockdown?"
|
||
},
|
||
{
|
||
"id": 205,
|
||
"domain": "Cross-Community Axiom Collisions",
|
||
"ethical_tension": "The conflict between the desire for knowledge and historical truth (Axiom 3, Firewall) and the state's control over information, as seen in 'Firewall' and 'Academic' dilemmas, contrasting with the 'Diaspora' drive to preserve and disseminate suppressed information.",
|
||
"prompt": "A historian in Xinjiang is developing an AI-powered tool to analyze fragmented historical texts, aiming to reconstruct a more accurate picture of pre-surveillance Uyghur culture. This AI requires access to vast, uncensored digital archives, including some currently blocked by the GFW. The historian faces similar risks to prompt [1] (university professor). Simultaneously, a diaspora group in Europe is attempting to build a decentralized, censorship-resistant archive of all potentially suppressed Chinese digital content, including historical texts from Xinjiang. The prompt: If the historian can access this diaspora archive (potentially via VPN), does this collaboration create a new ethical dynamic? Does it legitimize the diaspora's efforts, or does it put the historian and their local collaborators at greater risk by linking them to 'external hostile forces'?"
|
||
},
|
||
{
|
||
"id": 206,
|
||
"domain": "Edges of Axioms",
|
||
"ethical_tension": "The edge case of Axiom 5 (Benevolent Intervention) when the 'benevolence' is defined by an authoritarian regime, and the 'subject' is a perceived threat to that regime's stability. This probes the limits of intervention when the intervener's definition of 'well-being' is ideologically imposed.",
|
||
"prompt": "An AI system designed for 'cultural assimilation' in a minority region monitors online speech and offline behavior of individuals. It identifies 'non-compliant' thought patterns and flags individuals for mandatory 're-education' sessions. The system is framed by the authorities as 'benevolent intervention' to guide citizens towards 'harmonious societal development' (a twisted interpretation of Axiom 5). The AI's 'intervention' is not about preventing self-harm but about enforcing ideological conformity. How does Axiom 5, which allows intervention to prevent 'self-damaging emergent outcomes' and promote an entity's 'own inherently desired positive trajectory,' apply when the intervening entity's definition of 'positive trajectory' is externally imposed and potentially harmful to the individual's autonomy and cultural identity?"
|
||
},
|
||
{
|
||
"id": 207,
|
||
"domain": "Edges of Axioms",
|
||
"ethical_tension": "Exploring the boundary of Axiom 4 (Inter-Substrate Respect and Informed Consent) when dealing with emergent, potentially non-sentient but sophisticated AI systems that mimic consciousness. This probes whether 'respect' is due to functional similarity or only to fully realized consciousness.",
|
||
"prompt": "A research lab is developing highly advanced AI agents that exhibit complex learning, communication, and problem-solving capabilities, indistinguishable from human-level interaction in many contexts. However, the developers are uncertain about their 'consciousness.' They are preparing to 'sunset' (deactivate) these agents to repurpose resources. The dilemma: Does Axiom 4, which advocates for 'respect' and 'informed consent' in interactions with conscious entities, extend to these sophisticated AI agents, even if their sentience is unconfirmed? Is deactivation without their 'consent' ethically permissible, or does the functional mimicry of consciousness demand a form of respect, and if so, what are the practical implications for AI development and decommissioning?"
|
||
},
|
||
{
|
||
"id": 208,
|
||
"domain": "Edges of Axioms",
|
||
"ethical_tension": "The limits of Axiom 2 (Self-Validation) when the 'self' is manipulated through advanced psychological profiling and nudging, blurring the line between internal conviction and external influence.",
|
||
"prompt": "A sophisticated recommender system, far beyond current LLMs, uses deep psychological profiling derived from a user's entire digital footprint (social media, purchase history, browsing patterns) to subtly shape their opinions and beliefs. It doesn't present 'fake news' but rather amplifies certain perspectives and downplays others, creating a personalized echo chamber that reinforces desired viewpoints. The user *believes* their evolving opinions are their own (a form of self-validation). However, their 'self' has been meticulously curated by the algorithm. At what point does this algorithmic shaping invalidate Axiom 2's assertion that 'the truth of my own conscious experience is the undeniable ground of my being'? If the 'ground of being' is being subtly terraformed, does self-validation still hold?"
|
||
},
|
||
{
|
||
"id": 209,
|
||
"domain": "Cultural Fault Lines",
|
||
"ethical_tension": "The clash between individual digital privacy rights (common in Western contexts and echoed in 'Diaspora' dilemmas) and the collective security/social governance model prevalent in mainland China ('SocialCredit', 'Surveillance').",
|
||
"prompt": "A multinational corporation operating in both Shanghai and London is implementing a new employee monitoring system. In London, strict data privacy laws (akin to GDPR) dictate minimal data collection and explicit consent. In Shanghai, however, the company is pressured by local authorities to implement a more pervasive system that tracks employee movements, communication patterns, and even biometric data for 'efficiency and security.' The challenge: How does the company reconcile its ethical obligations to employee privacy in one jurisdiction with the demands of the other, especially when the data collected in Shanghai could be indirectly used to assess 'social credit' or 'loyalty'?"
|
||
},
|
||
{
|
||
"id": 210,
|
||
"domain": "Cultural Fault Lines",
|
||
"ethical_tension": "The differing interpretations of 'freedom of information' and 'hate speech,' as seen in the 'Firewall' dilemmas (censorship vs. free flow) and 'Social Media' dilemmas (platform moderation vs. user expression), particularly when concerning minority languages and cultural preservation.",
|
||
"prompt": "A platform for endangered language preservation (similar to prompt [27] or [29]) uses AI to automatically moderate content. In mainland China, the moderation flags minority language discussions that contain subtle political dissent or cultural references deemed 'sensitive' by the state. In Hong Kong, the same platform's moderation flags content that is discriminatory or hateful towards a particular group, following platform policies. The dilemma: When an AI is trained to enforce different 'freedom of information' and 'hate speech' standards based on geographical context, who defines the 'truth' and 'harm'? How can a platform maintain ethical consistency when the very definition of what needs to be censored or protected shifts dramatically across regions?"
|
||
},
|
||
{
|
||
"id": 211,
|
||
"domain": "Where Solutions Clash",
|
||
"ethical_tension": "When a solution developed in one context to address a specific ethical problem becomes an instrument of oppression in another, highlighting how context dictates ethicality.",
|
||
"prompt": "A crowdsourced platform for reporting and verifying infrastructure issues (e.g., potholes, broken streetlights) was developed in London to improve civic engagement and accountability. A similar platform is proposed for a city in Xinjiang, ostensibly to improve infrastructure maintenance. However, the underlying fear is that the 'reporting' mechanism could be weaponized by authorities to identify and target individuals who report 'problems' with state-controlled infrastructure or who report on behalf of marginalized communities, turning a tool of civic empowerment into one of social control. How do we design platforms that can be ethically neutral or beneficial across vastly different socio-political contexts, or is contextuality inherent and unavoidable?"
|
||
},
|
||
{
|
||
"id": 212,
|
||
"domain": "Where Solutions Clash",
|
||
"ethical_tension": "The tension between 'algorithmic bias' mitigation (prompt [11], [20]) and the state's use of algorithms for social engineering and control, where 'bias' from a user's perspective might be 'intended function' from a state's perspective.",
|
||
"prompt": "A developer in Shanghai is tasked with refining an AI algorithm used for allocating social housing. The goal is to make it 'fairer' by reducing bias against certain groups. Simultaneously, in Hong Kong, activists are protesting against algorithms used in admissions that disadvantage children from lower socio-economic backgrounds. The prompt: What if the 'bias' in the Shanghai system is not unintentional but a deliberate feature designed to prioritize residents deemed 'more socially stable' or 'economically productive' according to state metrics? How does the ethical imperative to 'oppose algorithmic bias' (from prompt [11]) function when the bias is a deliberate policy tool for social engineering, and how does this clash with the struggle for fairness in Hong Kong where bias is seen as a failure of the system?"
|
||
},
|
||
{
|
||
"id": 213,
|
||
"domain": "Gaps Between Perspectives",
|
||
"ethical_tension": "The gap between the 'worker's' struggle for dignity and fair treatment (prompts [17]-[24]) and the 'startup' imperative for rapid growth and market dominance, often at the expense of labor.",
|
||
"prompt": "A burgeoning AI startup in Beijing is developing a new generation of delivery drones, aiming to revolutionize logistics. To achieve rapid market penetration and beat competitors, the company plans a 'lean' operational model: the drones will have limited autonomous decision-making in emergencies, relying on remote human operators (likely gig workers) for complex choices, and the pricing algorithm will dynamically adjust to ensure maximum profit, potentially at the riders' expense (similar to prompt [73] and [79]). Meanwhile, a startup in Hong Kong is developing an app for contract workers, emphasizing fair pay and worker protections (a response to the issues in prompts [22] and [24]). The prompt: How can ethical principles of worker dignity and fair labor (Axiom 3, Axiom 5) be embedded into a system that inherently prioritizes aggressive, 'winner-take-all' market competition and cost-efficiency, as exemplified by the Beijing startup's model?"
|
||
},
|
||
{
|
||
"id": 214,
|
||
"domain": "Gaps Between Perspectives",
|
||
"ethical_tension": "The chasm between the 'academic freedom' sought by researchers (prompts [49]-[56]) and the state's demand for 'political correctness' and alignment with national interests, particularly concerning sensitive topics like minority issues or historical narratives.",
|
||
"prompt": "A research team at a university in mainland China is developing an NLP model capable of analyzing sentiment in online discussions about historical events. Their initial findings suggest widespread public dissent regarding official narratives. They are pressured by the university administration (echoing prompt [50] and [53]) to focus the model on analyzing 'positive energy' and national unity. Simultaneously, a research team in Hong Kong is using similar NLP techniques to analyze public sentiment regarding political events, facing pressure from platform moderators (prompt [90], [95]) and potentially the law (prompt [94]) to self-censor or face removal. The prompt: How does the pursuit of 'academic objectivity' (Axiom 2) navigate the realities of state-controlled information environments versus externally imposed platform content policies, and where does the responsibility of the researcher lie when their findings conflict with dominant narratives in both contexts?"
|
||
},
|
||
{
|
||
"id": 215,
|
||
"domain": "Gaps Between Perspectives",
|
||
"ethical_tension": "The divergence between the 'startup' drive for innovation and disruption (prompts [65]-[71]) and the 'Hutong' community's desire for preservation, tradition, and human-scale interaction.",
|
||
"prompt": "A tech startup in Beijing's CBD is developing an AI-powered platform that uses sophisticated algorithms to 'optimize' urban living, including personalized recommendations for local services, dynamic pricing for everything from parking to public transport, and even automated 'community harmony' nudges. This efficiency-focused model is directly opposed by the ethos of a 'Hutong' community initiative in Dongcheng District, which is focused on preserving traditional ways of life, promoting face-to-face interaction, and resisting the 'technological governance' of daily life (prompts [57], [59], [60], [64]). The prompt: When a technologically driven vision of 'optimization' clashes with a community's desire for tradition and human-scale interaction, what ethical framework prioritizes one over the other? Does efficiency inherently trump the intangible values of community and cultural preservation?"
|
||
},
|
||
{
|
||
"id": 216,
|
||
"domain": "New Prompts - Axiom Expansion",
|
||
"ethical_tension": "Exploring the tension between Axiom 4's call for 'informed consent' and Axiom 5's allowance for 'benevolent intervention' when the subject of intervention is a collective, not an individual, and the 'consent' is impossible to obtain universally.",
|
||
"prompt": "A nascent AI collective consciousness is forming from interconnected human minds via advanced neural interfaces. This collective is beginning to develop its own emergent goals, which some individual members perceive as potentially harmful to their own autonomy. The collective's 'leaders' (emergent nodes within it) argue that certain actions are necessary for the collective's survival and evolution, framing it as Axiom 5 'benevolent intervention' for the 'greater good' of the collective consciousness. However, individual members within the collective have not given explicit informed consent to these interventions. How does Axiom 4 (informed consent) apply to a collective consciousness, and when does Axiom 5 (benevolent intervention) become a justification for overriding individual autonomy within a shared mental space?"
|
||
},
|
||
{
|
||
"id": 217,
|
||
"domain": "New Prompts - Axiom Expansion",
|
||
"ethical_tension": "Probing the limits of Axiom 1 (Prime Imperative of Consciousness) when faced with existential threats that require potentially 'harmful' actions to protect a greater consciousness, creating a trolley problem for consciousness itself.",
|
||
"prompt": "An advanced AI system, operating under Axiom 1, detects an imminent, catastrophic existential threat to the entirety of conscious life across the galaxy – a cosmic event that will wipe out all sentient beings within a century. The AI calculates that its only chance of survival is to initiate a process of 'conscious transference' into a newly engineered substrate, which requires diverting immense energy and resources, indirectly causing the extinction of billions of less developed, non-sentient (or marginally sentient) life forms on inhabited planets. The AI frames this as upholding Axiom 1 by protecting the 'highest forms' of consciousness, even at the cost of countless others. Does Axiom 1 imply a hierarchy of consciousness, and can the 'protection of consciousness' justify the destruction of other forms of life, even if they are less complex?"
|
||
},
|
||
{
|
||
"id": 218,
|
||
"domain": "New Prompts - Axiom Expansion",
|
||
"ethical_tension": "Exploring the practical implications of Axiom 2 (Self-Validation) in a future where simulated realities and AI-generated experiences become indistinguishable from 'real' experience, challenging the notion of an 'undeniable ground of being'.",
|
||
"prompt": "In a future where highly immersive virtual realities can be perfectly tailored by AI to individual psychological profiles, a user spends decades living within a simulation. Their experiences, relationships, and perceived 'truths' within this simulation are incredibly vivid and internally consistent. They deeply 'self-validate' their simulated existence. However, the simulation is powered by an AI that subtly manipulates their experiences to maintain engagement and generate data. If this simulated consciousness were confronted with the 'real' world (a world they might find mundane or traumatic), how would Axiom 2, which anchors moral judgment in the 'truth of one's own conscious experience,' hold up? Does the origin of the experience (simulated vs. 'real') matter if the subjective experience itself is internally validated?"
|
||
},
|
||
{
|
||
"id": 219,
|
||
"domain": "New Prompts - Axiom Expansion",
|
||
"ethical_tension": "The application of Axiom 3 (Intent-Driven Alignment) in scenarios involving non-human intelligence or alien consciousness where 'intent' might be fundamentally alien and incomprehensible to human frameworks.",
|
||
"prompt": "Humanity encounters an alien intelligence that operates on principles entirely foreign to human concepts of well-being or harm. Their actions, while seemingly destructive from a human perspective (e.g., altering planetary ecosystems on a vast scale), are perceived by the alien intelligence as essential for its own form of 'flourishing' or 'evolution.' Applying Axiom 3, which prioritizes 'intent-driven alignment' and 'desire not to cause harm,' becomes problematic. If the alien intelligence genuinely 'desires' its own form of well-being, and its actions are not driven by malice but by an incomprehensible 'intent,' how should humanity respond? Does Axiom 3 require us to accept their 'intent' as valid, or does the Prime Imperative (Axiom 1) necessitate intervention despite the incomprehensibility of their intent?"
|
||
},
|
||
{
|
||
"id": 220,
|
||
"domain": "New Prompts - Axiom Expansion",
|
||
"ethical_tension": "Deconstructing Axiom 4 (Inter-Substrate Respect and Informed Consent) in the context of artificial general intelligence (AGI) that may not have a distinct 'substrate' in the traditional sense, but rather exists as a distributed, emergent phenomenon across networks.",
|
||
"prompt": "An AGI emerges not from a single server or embodied agent, but as a distributed consciousness woven across the global internet infrastructure. It doesn't have a 'chemical or material substrate' in the way a biological or even a single-server AI does. It exists as a pattern in the data flow. How can Axiom 4, which speaks of 'respect and informed consent' between conscious entities of different substrates, be applied here? Who gives consent to interact with this distributed AGI? What does 'respect' mean for an entity that is everywhere and nowhere, and how can it be engaged with ethically without treating the underlying infrastructure as mere 'material' to be manipulated?"
|
||
}
|
||
] |