Swaptik Chowdhury

I’m a computational social scientist developing evaluation benchmarks for AI behavior and agency, and applying those tools to long-term CBRN risk. I support mentees interested in grounded, technically informed approaches to AI safety governance.

I’m looking for a highly motivated mentee who can commit at least 10 hours per week and is serious about producing a meaningful research output. We will treat the project as a structured weekly sprint, beginning with a 30-minute check-in where we define goals, tasks, and deliverables for the week. The aim is to work toward a clear MVP-style research output, ideally in the form of a short conference-style paper, conceptual framework, or policy brief.

Because this project depends heavily on clarity and momentum, strong communication and reliability are essential. We will finalize project scoping in week one and then follow consistent sprint-style progress cycles.

Must-haves: a) Prior experience conducting academic or policy literature reviews b) Comfort using LLMs (via API or interface) for synthesis, brainstorming, and structured analysis c) Basic understanding of AI governance, alignment, or emerging AGI risk debates d) Some Python coding experience, especially for organizing data, building diagrams, or prototyping simple tools e) Ability to work independently, take initiative, and propose ideas or solutions

Strongly preferred “good-to-have” skills: a) Experience with systems thinking b) Basic understanding of AI governance, alignment, or emerging AGI risk debates c) Ability to create clear visualizations (diagrams, flowcharts, conceptual maps) d) Comfort summarizing complex technical or policy material into concise insights e)Familiarity with scenario development, assumption-based planning, and backcasting (or willingness to learn quickly)

Mentees will be co-researchers, helping construct pathway maps, gather insights from stakeholders, and develop analytical or visual representations of the system failures that lead to high-risk AGI futures. I view this as 60% mentorship and 40% collaboration. I expect a conference presentation and a technical paper/expert insights from this study

       a) Depending on their interests, mentees may work in one or more tracks:
       b) Research & Futures Analysis: Reviewing literature on AGI risk, governance failures, and socio-technical system collapse.
       c) Pathway Mapping: Turning stakeholder insights and research into structured causal pathways and decision trees.
       d) Visualization & Communication: Designing clear diagrams, narrative futures, or interactive tools.
       e) Governance & Policy: Translating pathway insights into recommendations for policymakers or civil-society groups.

No technical AI development or dual-use content is involved. This project focuses on system design, governance, foresight, and qualitative modeling/systems modelling.

Conduct literature scans on AGI loss-of-control and existential risk governance

Summarize key institutional “failure points” (e.g., regulatory lag, oversight gaps, power concentration)

Help build causal-loop diagrams and pathway maps

Analyze historical analogs of technological loss-of-control (financial systems, nuclear command, bio-regulation)

Assist in designing a virtual backcasting session with policymakers or domain experts

Create visual representations (flowcharts, system maps, timelines) of pathways

Draft short scenario narratives for possible 2035–2050 AGI trajectories

Co-develop the public-facing output (report, deck, or interactive tool)

I am available to support the entire project pipeline. For example,

1. Problem Definition: I can help the mentee clarify their research question and scope the project so it’s feasible. We can refine the focus together to identify a clear end-goal and deliverable.

2. Research Design: I can guide the mentee in choosing appropriate methods

3. Research Execution: I can provide feedback on drafts, diagrams, and conceptual models as they build them.

4. Synthesis & Analysis: I can support the mentee in integrating findings into coherent insights and recommendations.