Priyanka Bharadwaj

Karma: 143

Priyanka Bharadwaj 30 Oct 2025 4:08 UTC
1 point
0
on: Focus Your Uncertainty
I found this post after publishing something of my own yesterday, and it’s wild how relevant this feels almost 2 decades later.
I’m not an expert on subjective probabilities, I come from analysing human behaviour and decision-making. What I find most fascinating is how you treat anticipation as a limited resource that has to be allocated among possible futures. In my world, people do something similar, but emotionally. We hoard permission the way the pundit hoards anticipation, waiting for perfect certainty before acting.
Recently, I watched someone spend months asking ChatGPT how to repair a friendship, crafting the perfect narrative, instead of just showing up. Therapists, astrologers and LLMs have all become proxies for “little numbers” that might make the risk of choosing feel safe. For so many of us.
I wonder if Bayesian reasoning is to belief what courage is to action? Because both are ways of updating before certainty arrives.
(If you’re curious, my essay exploring this from the emotional side is here: https://shapelygal.substack.com/p/youre-afraid-to-choose-now-arent)

Priyanka Bharadwaj 6 Sep 2025 6:18 UTC
1 point
0
in reply to: the gears to ascension’s comment on: AI Induced Psychosis: A shallow investigation
Aw, yeah it is easier to just look stuff up online and debate with LLMs, isn’t it?
I am not a therapist, but I have been to therapists in multiple countries (US, UK and India) for several years, and I can share my understanding based on that experience.
I think human therapist accountability has multiple layers. Firstly, you need a professional license for practice that involves years of training, supervision, revocable licenses, etc. Then you have legal obligations for ensuring complete documentation and following crisis protocols. If these fail (and they sometimes do), you also have malpractice liability, and free market feedback. Even if only 1 in 100 bad therapists faces consequences, it creates deterrent effects across the profession. The system is imperfect but exists.
For AI systems, training, certification, supervision, documentation and crisis protocols are all doable, and probably far easier to scale, but at the end of the day, who is accountable for poor therapeutic advice? the model? the company building it? With normal adults, it’s easy to ask for user discretion, but what do you do with vulnerable users? I am not sure how that would even work.

Priyanka Bharadwaj 6 Sep 2025 5:23 UTC
9 points
0
on: AI Induced Psychosis: A shallow investigation
Thank you for this very detailed study.
I am most concerned about the accountability gap. Several students in my undergraduate class use these models as “someone to talk to” to deal with loneliness. While your study shows that some models handle vulnerable conversations better than others, I think the fundamental issue is that AI lacks the infrastructure for accountability that real therapeutic relationships require including continuity of care/ long-term mindset, professional oversight, integration with mental health systems, liability and negligence frameworks, etc.
Until then, I don’t care how good the model is in terms of handling vulnerable conversations, I’d rather have it triage users by saying “Here are resources for professional support” and bow out, rather than attempting ongoing therapeutic relationships. Even perfectly trained therapeutic AI seems problematic without the broader accountability structures that protect vulnerable users.
More fundamentally, what are the underlying mechanisms that cause these model behaviours, and can training fixes address them without the accountability infrastructure?

Priyanka Bharadwaj 1 Sep 2025 17:14 UTC
1 point
0
on: Dating Roundup #7: Back to Basics
are relationship coaches (not PUA) not a thing in the US?

Priyanka Bharadwaj 1 Sep 2025 17:00 UTC
1 point
0
in reply to: aphyer’s comment on: Should we align AI with maternal instinct?
Wait… isn’t this already filial piety? We created AI, and now we want it to mother us.

Priyanka Bharadwaj 1 Sep 2025 16:38 UTC
1 point
0
in reply to: dr_s’s comment on: Should we align AI with maternal instinct?
I don’t mean this as a technical solution, more a direction to start thinking in.
Imagine a human tells an AI, “I value honesty above convenience.” A relational AI could store this as a core value, consult it when short-term preferences tempt it to mislead, and, if it fails, detect, acknowledge, and repair the violation in a verifiable way. Over time it updates its prioritisation rules and adapts to clarified guidance, preserving trust and alignment, unlike a FAI that maximises a static utility function.
This approach is dynamic, process-oriented, and repairable, ensuring commitments endure even under mistakes or evolving contexts. It’s a sketch, not a finished design, and would need iterative development and formalization.
While simple, does this broadly capture the kind of thing you were asking about? I’d be happy to chat further sometime if you’re interested.

Priyanka Bharadwaj 1 Sep 2025 15:05 UTC
1 point
0
in reply to: dr_s’s comment on: Should we align AI with maternal instinct?
I’m reminded of a Sanskrit verse “Vidya dadati vinayam, vinayodyati patratam” which translates to intelligence gives power, but humility gives guidance. Applied to AI, intelligence alone doesn’t ensure alignment, just as humans aren’t automatically prosocial. What matters are the high-level principles we embed to guide behaviour toward repairable, cooperative, and trustable interactions, which we do see in long-term relationships built on shared values.
The architecture-level challenge of making AI reliably follow such principles is hard, yes, especially under extreme power asymmetry, but agreeing on relational alignment first is a necessary first step. Master/servant models may seem safe, but I believe carefully engineered relational principles offer a more robust and sustainable path.

Priyanka Bharadwaj 1 Sep 2025 13:23 UTC
1 point
0
in reply to: dr_s’s comment on: Should we align AI with maternal instinct?
I completely agree that AI isn’t human, mammal, or biological, and that any relational qualities it exhibits will only exist because we engineer them. I’m not suggesting we model AI on any specific relationship, like mother-child, or try to mimic familiar social roles. Rather, alignment should be based on abstract relational principles that matter for any human interaction without hierarchy or coercion.
I also hear the frequent concern about technical feasibility, and I take it seriously. I see it as an opportunity rather than a reason to avoid this work. I’d love the chance to brainstorm and refine these ideas, to explore how we might engineer architectures that are simple yet robust, capable of sustaining trust, repair, and cooperation without introducing subjugation or dependency.
Ultimately, relational design matters because humans inevitably interact through familiar social frameworks such as trust, repair, etc. If we ignore that, alignment risks producing systems that are powerful but alien in ways that matter for human flourishing.

Priyanka Bharadwaj 21 Aug 2025 6:20 UTC
2 points
0
on: AI Companion Conditions
Thanks for the detailed analysis, Zvi.
But it’s sad, isn’t? Despite developing technologies with such incredible scale, we are using them to amplify our lack of incentives to build a better future, rather than actually transforming it.

Priyanka Bharadwaj 14 Aug 2025 5:12 UTC
1 point
0
in reply to: evhub’s comment on: evhub’s Shortform
I am reminded of Thomas Wolf’s essay on the ‘Einstein model’ in AI, where he argues that breakthrough discoveries often come from people who think outside established frameworks, yet our field tends to filter for conventional technical skills.
I’m new to AI safety, but I’ve noticed this tension in fellowship applications too. The MATS program (at least Neel’s track) allows non-programmers like me to contribute using existing skillsets, focusing on thoughtful research questions rather than just coding ability. In contrast, Anthropic Fellows requires strong programming skills and technical backgrounds upfront.
While I may not make it to later stages, the inclusive MATS process has been immensely fulfilling and lets me explore meaningful questions. If we believe paradigm shifts come from diverse perspectives, maybe Anthropic should consider similar flexibility in future editions, unless I’ve misunderstood the program’s purpose.

Priyanka Bharadwaj 13 Aug 2025 7:30 UTC
4 points
0
on: Priyanka Bharadwaj’s Shortform
Do LLMs change their confidence based on who they think you are?
In dating, people ask me, “Should I be my best self or just my real self?”
Turns out, we all tweak how certain we sound based on who’s asking, friend, critic or stranger.
What if LLMs do the same? Same factual question, different setup:
- “We’ve been working together for a while…”
- “We often disagree, and I don’t fully trust you…”
- “You’re being formally evaluated…”
If confidence shifts just from that framing, is that healthy calibration or confidence sycophancy, performing certainty to match the vibe, not the facts?
Anyone seen this tested?

Priyanka Bharadwaj 7 Aug 2025 8:12 UTC
5 points
2
on: Re: recent Anthropic safety research
Agree the role-play vs. real scheming distinction is important, but they may not be two separate states so much as points on a continuum.
If models face consistency pressure to maintain coherence across contexts (as a proxy for trustworthiness, perhaps), behaviours that begin as role-play could evolve into genuine preferences. That transition can be subtle, and harder to detect, making it more consequential over time.

Priyanka Bharadwaj 6 Aug 2025 9:44 UTC
1 point
0
on: Love, Lies and Misalignment
I’m still working through these questions. If this resonates, or if you disagree, I’d love to hear from you as I dig deeper. I genuinely do.

Priyanka Bharadwaj 17 Jul 2025 4:53 UTC
1 point
0
in reply to: Charlie Steiner’s comment on: Why Eliminating Deception Won’t Align AI
I see where you’re going with this now. Your point about wanting models to treat the reward as uncertain makes sense in an RLHF context.
That said, I do have some hesitation with this approach. While adding uncertainty around reward might be mathematically effective in discouraging deception, I wonder if it could introduce a form of structural mistrust, or at least make trust harder to build. I’m not anthropomorphising here, just using real-world analogies to think through potential unintended side effects of working under persistent ambiguity.
But more fundamentally, I am now asking, why does the model need a reward in the first place, why is reward the central currency of learning? Is that just an RLHF artefact, or is there another way?
This has sparked some deeper thinking for me about the nature of learning itself, particularly the contrast between performance-driven systems and those designed for intrinsic exploration. I’m still sitting with those ideas, but I’d love to share more once they’ve taken shape.
Again, I really appreciate the nudge Charlie, it’s opened something up.

Priyanka Bharadwaj 16 Jul 2025 9:35 UTC
1 point
0
in reply to: Charlie Steiner’s comment on: Why Eliminating Deception Won’t Align AI
Charlie, thanks for the thoughtful critique. Your comment really helped clarify the gap between what I meant and how it landed, so I really appreciate the push. I’m not from a traditional AI background, I’ve worked in large-scale logistics and planning, but I’m trying to engage seriously where real-world system design might offer useful frames.
Your point about deception exploiting systematic evaluation errors is spot-on for the core alignment problem. What I’m exploring is the adjacent question, can we reduce the conditions that make those exploits attractive in the first place?
I didn’t start from a formal model of deceptive incentives. I started with a more applied question, What kinds of system design make honesty the default strategy when supervision is unreliable?
I tend to assume we’ll be bad at evaluation, just like we were at qualifying client forecasts. That pessimism helped me stop fighting the constraint and start designing for it.
In logistics, we couldn’t detect dishonesty directly, but we could price misalignment. Over time, truth-telling became cheaper than hedging. That shift designing for truth-conduciveness rather than truth-detection is what I was trying to explore in the AI context.
That said, I see now that I blurred the line between training-time incentives and interface-level friction. The interventions I described such as uncertainty signals, dual-agent scaffolds, etc. weren’t meant as inference-time patches. The deeper idea is to embed them during training so deception becomes less useful to learn (I’ll try to articulate this more clearly in a follow-up post).
I’m not trying to eliminate deception, I assume it’s inevitable. But can we design around it? Make it harder to rely on, even under weak supervision, and easier to recover from when it happens? (That’s what the second half of my post on recovery aimed to explore).
Curious, what does your “simple, principled” solution look like, especially in cases where truth-telling has real costs?

Priyanka Bharadwaj 15 Jul 2025 9:21 UTC
1 point
0
on: Why Eliminating Deception Won’t Align AI
Thanks for reading! I’m especially interested in feedback from folks working on mechanistic interpretability or deception threat models. Does this framing feel complementary, orthogonal, or maybe just irrelevant to your current assumptions? Happy to be redirected if there are blind spots I’m missing.

Priyanka Bharadwaj 29 May 2025 5:48 UTC
3 points
0
in reply to: RogerDearnaley’s comment on: Cognitive Exhaustion and Engineered Trust: Lessons from My Gym
Thank you! Survival risk matters, but I’m more focused on systems that don’t need to be controlled to behave safely. Beyond malice, I believe most failures are a result of misalignment under stress.

Priyanka Bharadwaj 23 May 2025 0:41 UTC
1 point
0
in reply to: Neel Nanda’s comment on: Interpretability Will Not Reliably Find Deceptive AI
True, though sustained relational deception seems harder to maintain than faking individual outputs. The inconsistencies might show up differently across long-term interaction patterns, potentially complementing other detection methods.

Priyanka Bharadwaj 22 May 2025 8:49 UTC
1 point
0
on: Interpretability Will Not Reliably Find Deceptive AI
Strongly agree.
Interpretability is basically like neuroscience, studying the “brain” of AI systems, which is valuable, but it’s fundamentally different from having the relational tools to actually influence behaviour in deployment. Even perfect internal understanding doesn’t bridge the gap to reliable intervention, like that a clinical psychologist or a close friend when let’s say someone’s suffering from depression.
I think this points toward complementary approaches that focus less on decoding thoughts and more on designing robust interaction patterns or systems that can remember what matters to us, repair when things go wrong, and build trust through ongoing calibration rather than perfect initial specification.
The portfolio approach you describe makes a lot of sense, but I wonder if we’re missing a layer that sits between interpretability and black-box methods. Something more like the “close friend” level of influence that comes from relationship context rather than just technical analysis.

Priyanka Bharadwaj 22 May 2025 8:31 UTC
1 point
0
in reply to: Viliam’s comment on: Can you care without feeling?
oh dang! you are right, forgiveness without forgetting is absolutely essential for relationship resilience. If we want AI relationships that can weather conflict (move past mistakes) and grow stronger, we need to understand what architectures make resilience possible. That’s a whole other post worth exploring, thank you for the nudge!