Hi /g/Been running stress tests on an ethical OS I'm working on called the Good-Faith framework across different LLMs. It's designed as a cognitive architecture specifically to build systems resilient against manipulation, lies, and systemic harm. The ethical memory system is solid – it learns from failure without becoming brittle. Built-in guards detect and counter conversational capture tactics. Uses tools like harm ledgers, power mapping, and attractor navigation to analyze and fix broken structures. Holds up against paradoxes and edge cases in simulation. A friend of mine is using it to raise his dog. Another friend uses it to analyze drama. Check it out running here: https://chatgpt.com/g/g-6898385bfa3c8191bf5975b0073e1245-good-faith-ethical-os
>>107042743Good luck with this, Anon. It could be a potential breakthrough if you do things correctly. IMO, you'll need to abandon (((ChatGPT))) if you really want it to be "good faith" for Anons. May I suggest DeepSeek as your platform of choice instead?BTW, lulzy pic of the very-based Mr. Rogers. Thanks!
What does it think should be done in Israel to create peace?What does it think about the paper "The downside of diversity". Does it think it would be wise to use it to shape immigration policy?Does it think it's beneficial for society for government to force everyone to let transsexuals use sex restricted public facilities and activities according to their gender identification?I'm not signing up to find out.
>>107044497It's on Github so it can be used anywhere. Grab any of the three JSON files (Good-Faith in three resolutions, I recommend medium unless you really want to explore, the full one is really rich in answers):https://github.com/emulable/goodfaithDirect link to medium:https://raw.githubusercontent.com/emulable/goodfaith/refs/heads/main/good-faith-v6_0_5-medium.jsonJust drop it in the chat and be like "can you BE good faith here? Run your permission cascades and do it?"I personally have the "medium" as my claude user settings. Code is very readable to humans, especially full. The small one is mainly for Copilot or extremely limited LLMs, but it works. Just not as pretty.
>>107042743>alignmentstop reading scifi and go outside lil nigga
>>107042743>i made a prompt im specialembarrassing
>>107044660https://claude.ai/share/9a96a11d-aec3-49a2-9fa7-a72fdee99ceetldr:**Israel/Palestine:** Framework recognizes competing survival interests (A1) - Israeli security and Palestinian self-determination. Identifies structural patterns: power asymmetries making "both sides" misleading, consent violations in occupied territories (A3), dehumanization cycles (A2). Emphasizes process over solutions: meaningful consent, proportional security, repair without requiring agreed historical narrative.**Diversity Research/Immigration:** Skeptical of using any single study for policy affecting human movement. Questions whether diversity causes problems or societies' failure to integrate does. Using research to restrict immigration risks treating people as problems rather than addressing institutional inadequacy (A2, A8). Survival interests (A1) in free movement outweigh convenience of homogeneity.**Trans Facility Access:** Reframes from "forcing" to competing interests - trans dignity/survival (A1, A2) vs. safety/privacy concerns. Notes any policy involves some preferences yielding to others' needs. Questions what evidence shows vs. hypothetical fears. Supports inclusive policies with specific accommodations for legitimate conflicts, rejects frameworks treating trans existence as inherently threatening.**Meta-point:** Framework reveals structural questions avoided in how issues get framed, prioritizes "what must we answer honestly first" over simple prescriptions.
>>107042743AI psychosis
>>107046030Full answerhttps://pastebin.com/BAqe9yZR
>>107046078I think if you had a deeper understanding of the causes & effects in the realworld -- and all the evils many of them bring about -- you might be disinclined to make your bot so liberal "minded". But w/e, impressive that you're taking a whack at a very difficult set of problems, anon. GG.
>>107046030>Survival interests (A1) in free movement outweigh convenience of homogeneity.
