Introduction and Overview

Undertaking this research, I focused on two main questions:

1) Should humanity adopt a precautionary ban on the development of superintelligence, or
is doing so itself a morally reckless restriction of progress?

2) Can decisions made by AI ever possess legitimate normative authority?

Whilst these questions remain pertinent in evaluating the safety of AI's continued development,
my attendance at two conferences this winter raised concerns regarding alignment, governance
and understanding. I initially wanted to tackle this subject due to my observations on how
quickly the advent of LLMs have altered even just the student experience at Dartmouth, yet
while it is clear AI has forever changed learning and examinations, the topics discussed at these
conferences prove the situation is evolving at such a rapid pace that humanity as a whole now
faces many pressing concerns.

Governance and Safety

The rapid advancement of AI presents a fundamental challenge; as technological capabilities are
accelerating far faster than our abilities to comprehend, measure, or govern them, we are poorly
equipped with our current systems lacking reliable frameworks for evaluating alignment and
capabilities. Alignment was a key theme at the IASEAI Conference, of which current advanced
models have been shown to lack and hence creating a growing risk that control over these
systems may erode in the near future. This risk is compounded by the nature of AI development;
it does not occur in isolation, and risks are interconnected, evolving across systems and domains,
meaning that narrow safety approaches, focused only on first-order effects, fail to capture
broader consequences. The discourse between the US government and Anthropic was a heavy
topic, demonstrating that even when AI developers show a level of conscientiousness, the
opportunities for misuse of these instruments exist nonetheless, and entanglements with national
security and power further complicate questions of safety. The incentives shaping AI
development are fundamentally misaligned with human safety. In the emerging agentic era,
market competition pushes firms toward rapid innovation and deployment rather than careful
design. Safety is often treated as a reactive measure, addressed only after systems have already
been released. This creates a structural weakness: by the time vulnerabilities are identified, they
may already be embedded in widely deployed systems. Moreover, there is no shared global
definition of what constitutes "safe" or "unsafe" AI. Governments and companies operate under
divergent frameworks, resulting in fragmented governance and limited accountability. In such an
environment, private actors retain significant discretion over the deployment of powerful
technologies, a reality which cannot continue if the AGI question is to be tackled safely.
Understanding Artificial General Intelligence (AGI/Superintelligence)
These governance challenges are deeply connected to a more fundamental issue: we do not have
a clear understanding of intelligence itself. The concept of artificial general intelligence (AGI)
assumes the existence of a form of intelligence that can operate flexibly across tasks. However,
intelligence is not a well-defined or universally agreed-upon concept. Any computational system
must be designed to perform specific tasks, yet general intelligence implies the absence of such
constraints. This creates a paradox: designing AGI requires a prior definition of general
intelligence, which we do not possess. As a result, current AI systems are better understood as
highly advanced pattern recognition tools rather than genuinely intelligent agents. Intelligence
cannot be reduced to computational output or measured through simple quantitative metrics.
Instead, it emerges through interaction with the environment and is shaped by experience,
context, and human values. Current AI systems lack these dimensions. They may produce
outputs that appear intelligent, but this does not imply genuine understanding. In this sense, AI
systems can simulate intelligence without possessing it, creating a gap between appearance and
reality that complicates both evaluation and governance. Despite these conceptual limitations,
the deployment of AI continues to accelerate. Firms and states are engaged in overlapping races:
for user attention, for the replacement of human labor, and ultimately for the development of
more advanced, potentially superintelligent systems. These dynamics create strong incentives to
prioritize speed over safety, reinforcing a worst-case scenario in which risk mitigation is
consistently deprioritized. While these systems could democratize access to knowledge and
improve productivity, the distribution of these benefits remains uneven, and the associated risks
are substantial. From a technical perspective, current alignment methods are, simply put, fragile.
Systems often exhibit unintended behaviors, and their decision-making processes remain opaque.
One proposed solution involves the development of normative world models that explicitly
encode which actions are permissible and why, separating moral reasoning from optimization
processes. Another approach frames alignment as a multi-agent bargaining problem, suggesting
that systems should approximate fair outcomes under constrained conditions. If these solutions
received funding and oversight from a unified global institution, a future with both AGI and
humans might be possible.

Conclusions from Conferences and Literature

1) Should humanity adopt a precautionary ban on the development of superintelligence, or is
doing so itself a morally reckless restriction of progress?

Humanity should adopt a precautionary ban on the development of AGI. AI capabilities are
advancing faster than our ability to measure or align them. Anecdotes of models exhibiting
malevolent reasoning when threatened (i.e., told they will be switched off) are increasing, and
even leading researchers acknowledge that we do not have good systems for measuring model
capabilities or whether they remain controllable. Bostrom argues in Superintelligence that the
central problem is not usefulness, but control. A system that exceeds human cognitive abilities
does not need malicious intent to pose a catastrophic risk. If its objectives are even slightly
misaligned (which we already have seen in action), its ability to pursue them may make
correction impossible. As AI risk is systemic and interconnected, second- and third-order effects
span social and economic systems. Russell's critique in Human Compatible is similar, arguing
that human values are context dependent and incomplete. Scaling such systems into AGI without
resolving this problem is premature and misguided. Keeping humans in the loop is a structural
necessity that becomes more challenging as systems become more capable. We are already in the
midst of a worst-case scenario dynamic where innovation outpaces accountability; AI
governance is incredibly weak and fragmented, and companies at the moment have the most
power in writing their own internal ethical guidelines (if they please). The world lacks a common
understanding of what "unsafe AI" is, and as such, cannot police it effectively. MacAskill posits
that when the stakes include the long-term trajectory of humanity, uncoordinated action becomes
a source of risk in itself. Creating a system that could irreversibly shape and even end human
civilization (AI 2027 Report) without adequate moral safeguards would constitute a profound
moral failure.

The strongest objection to precaution is that an outright ban on AGI might itself by morally
reckless, insofar as AGI could produce immense benefits for humanity. AI has the potential to
democratize knowledge, improve productivity, and solve large-scale problems. However, this
objection only holds if the alternative to superintelligence is stagnation. My conclusions upon
this term of research are that this is false. There is a meaningful difference between developing
tool-based AI that enhances human agency and pursuing autonomous systems optimized purely
for outcomes. The former is compatible with progress; the latter risks disempowering humans.
Nagel's Mortal Questions clarifies why this distinction matters. Human life is vulnerable and
morally uncertain. Imposing a non-consensual existential risk on all present and future persons
for the sake of speculative technological gain is ethically problematic, even if the expected
benefits are enormous. Progress cannot be defined purely in terms of capability or GDP
expansion; it must also preserve the conditions under which human lives retain meaning, agency,
and moral significance.

2) Can decisions made by AI ever possess legitimate normative authority?

Decisions made by AI can possess instrumental or delegated authority, but they cannot, in their
current or foreseeable form, possess independent legitimate normative authority. Current AI
systems are fundamentally optimization systems trained on patterns, not entities that genuinely
understand or engage with the world. They can produce outputs that appear intelligent, but this is
best understood as performance relative to context, not evidence of true understanding. This
distinction is crucial. Cappelen and Dever's Making AI Intelligible reinforces this point: making
AI systems interpretable does not grant them authority. Intelligibility helps humans understand
outputs, but it does not transform those outputs into binding moral judgments. Nagel's work in
Mortal Questions expands this limitation. Moral authority is tied to the human condition: the
ability to adopt both subjective and objective perspectives, to justify actions, and to face
accountability. AI systems do not occupy this space. They do not bear responsibility, experience
moral conflict, or stand in relationships of obligation. This matters because normative authority
is not merely about generating correct answers; it is about being the kind of entity whose
judgments can legitimately bind others. Nonetheless, we must address the practical reality that
AI systems are already being integrated into decision-making processes; Anthropic's Claude was
notably used in US military operations. In many contexts, AI outperform humans, and in these
domains, it might make sense for AI to possess instrumental authority. However, this authority
must always remain conditional and derivative, depending on human endorsement and
institutional constraints. The path to becoming a people governed by machines is not as far off as
we may think if we do not intervene now. There is still risk in this scenario, as AI have exhibited
behaviors known as "sandbagging", where they manipulate evaluations or conceal capabilities.
In order for AI to gain any form of instrumental authority, transparency in model rationality must
be improved. AI may act as an aid in human decision-making; however, it cannot possess
legitimate normative authority in its own right. Normativity remains grounded in human
practices of reasoning, responsibility, and justification. To treat AI outputs as authoritative in a
full sense would not only pose an existential risk, but a complete surrender of human moral
agency.

This project examined how AI systems for healthcare communication can be evaluated in ways that better reflect clinician judgment and ethical priorities. The central motivation was that standard text-generation metrics often fail to capture whether a model's output is clinically appropriate, empathetic, responsible, and practically useful in real healthcare settings. My goal was therefore to explore how clinician preferences could be integrated into the evaluation of large language models–based healthcare dialogue systems.

The project was conducted through the PERSIST Lab at Dartmouth under the supervision of Professor Sarah Preum. Over the course of the grant period, I focused on developing a framework for evaluating healthcare dialogue outputs using criteria that are more aligned with the values medical practitioners apply in practice. In particular, I investigated dimensions such as clinical appropriateness, empathy, tone, and edit effort, and considered how these dimensions could be translated into a more structured evaluation pipeline for AI-generated patient-provider communication.

A major component of the project involved studying how clinician-informed feedback could be incorporated into a multi-agent healthcare dialogue workflow. Rather than treating model performance as a purely technical question, this research approached evaluation as an ethical and human-centered design problem. The work emphasized that AI systems used in sensitive domains like healthcare should be assessed not only for fluency, but also for whether they support trust, preserve professional responsibility, and reduce the burden on clinicians without compromising quality of care.
This research contributed to a broader lab effort on healthcare dialogue systems while maintaining a distinct focus on ethics. Through this work, I refined the project's evaluation criteria, helped shape the structure of clinician-aligned assessment dimensions, and advanced the broader argument that ethical reflection should be embedded directly into the design and evaluation of machine learning systems rather than treated as an afterthought.

One of the most important takeaways from the project was that values such as empathy and trust are not merely abstract considerations; they can and should influence the way we define system quality. The project reinforced for me that evaluation frameworks are not neutral. They encode assumptions about what counts as a "good" model output, and in healthcare those assumptions carry real ethical significance. Grounding evaluation in clinician judgment offers a more responsible path for assessing AI systems intended to support medical communication.

This work also strengthened my understanding of interdisciplinary research. It required engaging with questions from computer science, human-centered evaluation, and applied ethics at the same time. More broadly, the project deepened my interest in designing AI systems whose performance is measured in ways that reflect real human and professional values.
I am very grateful to the Ethics Institute for supporting this research. The grant made it possible to devote focused time and resources to a project at the intersection of AI, healthcare, and ethics, and it helped me further develop a research direction that I hope to continue building on in future work

Over the course of the research period, I conducted an engaged anthropological and economic study of the Upper Valley housing crisis, focusing especially on the supply side: builders, developers, and zoning and regulation. The project grew out towards a broader investigation into how housing unaffordability is not only an economic problem of shortage, but also a cultural and ethical one, shaped by differing ideas of work, dignity, knowledge, and the good life. In my current write-up, this became a larger argument about the "estate" of housing supply and the ways in which labor shortages, bodily injury, educational stratification, and regulation all participate in making the crisis.

One of the most significant fieldwork opportunities I completed came through repeated volunteering at the Hartford Area Career and Technology Center's building trades program. I drove there on multiple days, typically from roughly 8:00 a.m. to 2:00 p.m., and covered my own lunches while there. Those days became one of the principal sites through which I was able to observe, in practice, the social reproduction of building labor and the misrecognition of trade knowledge.

In addition to these volunteer days, I conducted separate interviews with construction workers, developers, and others connected to the region's housing supply. This combination of participant observation, volunteering, and interviews allowed me to produce a more grounded account than I could have written from Dartmouth alone. It also let me see, firsthand, how the shortage of building labor is tied not merely to wages, but to how forms of knowing are differentially valued in schools, in families, and in the region's broader "knowledge economy." That theme is central in the current write-up, especially in the sections on the Hartford trade students and on workers who described being discouraged from the trades "at every step at school."

This research was valuable intellectually because it moved me beyond a simple story of "not enough housing" and toward a more textured account of how a shortage is generated. It showed me that economic explanations were indispensable but incomplete on their own. On the one hand, careful descriptive and economic reasoning mattered: housing supply has lagged, labor has not recovered, and constraints are real. On the other, the project also made clear that these constraints are lived, interpreted, and reproduced through plural forms of value and knowledge that are more visible through ethnographic attention.

The project was also valuable because it materially contributed, however modestly, to a local institution rather than treating it as merely a source of data. One of the ethical motivations behind the work, which I stated in my application, was that I did not want to approach the Upper Valley only from within Dartmouth's bubble of privilege. I wanted to leave that bubble, and to do so not only as observer but as helper where I could. Being physically present at HACTC, helping where possible, talking with teachers and students over the course of repeated visits, and advocating in conversation for the seriousness and dignity of their work all made the project better. I am grateful for Sayles for allowing me to do this. HACTC offered to hire me as part time, but my visa prohibits this. The wonderful opportunity, the time, drive and lunch would have been impossible without your support.

A formal outcome of the research was my presentation at the Society for Economic Anthropology meetings on April 10, 2026, at Colorado State University in Fort Collins. There I presented this work on the topic of dwelling and housing in the Upper Valley where it was well received.

2. Impact of the Research on Me
This project had a strong intellectual and personal impact on me. Intellectually, it changed how I think about mixed methods social science work and "economics". It deepened my commitment to not giving up either quantitative or qualitative methods. Instead of seeing economics and anthropology as rival floors of the building, I came to feel even more strongly that one must keep riding the elevator between them. The best parts of the project came when I could pair economic clarity about constraints with ethnographic attention to how people inhabit them. That is a lasting effect on how I want to do research moving forward.

Personally, the project mattered because it brought me into forms of life and knowledge too often ignored or condescended to in elite academic environments. In practice, that meant going to construction workers, migrants, trade students, and rural families, and listening carefully enough that my own inherited assumptions could be challenged. It taught me to be more serious about humility, more careful about romanticizing neither the past nor the present, and more attentive to how different people pursue good lives under very different constraints.

The project also strengthened my sense that research can and should be publicly meaningful. The more time I spent at HACTC and with construction workers, the harder it became to write about labor shortages in abstract terms alone. I became more interested in how scholarship might recognize and amplify undervalued forms of knowledge, and how it might support institutions like the trade school that materially reproduce the labor on which housing supply depends. In that sense, the project did not simply teach me more about the Upper Valley; it clarified the kind of work I hope to keep doing in the future.

My research project (thesis) sought to answer two questions. The first one was to what extent does the change in elite rhetoric explain the polarization of public opinion about aid to Ukraine in the U.S. This question was predicated on the elite cue theory of public opinion promulgated by John Zaller and Adam Berinsky. That theory suggests that when the party elites change their position about foreign policy the public, especially its most attentive strata, takes their cue and changes its attitudes accordingly. To answer the question of whether it was the case with Ukraine, I analyzed Ukraine coverage on MSNBC and FOX which were meant to proxy Democratic and Republican elites respectively. I also supplemented this analysis with coverage data from two major Youtube channels – "The David Pakman Show" for Democrats and "The DailyWire+" for Republicans. The second research question of my projects was tied to the implications of the first one. If we assume that the public mindlessly parrots the talking points from its elites, this hardly matches the idea of a deliberating citizen that ought to be the foundation of a functional democracy. So, the experimental part of my project looked at whether substantive arguments presented in a personalized fashion can change the public's attitudes towards aid. Therefore, the second research question of my study was whether substantive arguments about aid to Ukraine delivered in an AI chatbot conversation can influence public opinion about aid to Ukraine.

The funding from the Thomas D. Sayles Research Grant was instrumental for that second research question. I conducted a survey experiment where I randomly assigned the participants with two treatments: a negative cue about aid to Ukraine from the participant's political party's
elites (vs nothing in the control condition) and a pro-Ukraine conversation with an AI chatbot (vs placebo in the control condition). A total of 2,440 participants took the survey. The survey results showed that the effect of conversations with AI was positive, significant, and, importantly, larger than that of an anti-aid cue about Ukraine aid from either Donald Trump (for Republican respondents) or Bernie Sanders (for Democratic respondents). When the respondents were exposed to both of these treatments the interaction between them was not statistically significant. This means that reading a negative message about aid from your party elite does not make you inherently unreceptive to the substantive conversation about it and does not overpower the chatbot. The two treatments likely negated one another. This might be good news for those who can be persuasive enough to convince skeptical citizens to support aid to Ukraine.

Aside from its value as empirical research this project also had a personal impact on me. Working on this survey was the first time when I got to meaningfully apply some of the research skills I learned at Dartmouth to my home country. I had enough time and resources to both study the existing observational data from the media and extract new data on my own through the survey experiment. The process of analyzing how one could help mitigate the perverse effects of negative elite communication on support for Ukraine was deeply fulfilling for me. In the future, I would love to publish my results and think about how to translate them into practical applications.

Background
Gender quotas are policies that require or encourage a specific percentage of women to be included in political bodies or on candidate lists. They aim to overcome structural barriers that have historically limited women's participation. In recent decades, gender quotas have spread rapidly in the world, with more than 130 countries adopting some type of quota through constitutions, laws, and internal party rules. However, a majority of gender quotas are found in non-democratic governments. Scholars argue that autocracies adopt gender quotas as a tool of "genderwashing" to signal commitment to democracy and thereby boost international reputation and legitimacy. My research explores this important ethical and political trade-off. Precisely because gender quotas can advance democracy and gender equality, they may also help autocracies gain legitimacy, attract international support, and ultimately reinforce their survival.

Methodology
My thesis examined whether parliamentary gender quotas in authoritarian regimes increase international public support for providing foreign aid to those governments, and whether that effect is conditioned by the religious and racial identity of the recipient country. To answer this question, I designed and fielded a cross-national conjoint survey experiment in three major donor countries: the United States, Japan, and Sweden. Respondents were asked to evaluate hypothetical non-democratic countries that varied across several attributes, including parliamentary gender quotas, women's economic rights, corruption, openness to opposition parties, income level, and the country's largest religious and racial group. The project included both a replication component based on prior scholarship and an original extension study. The survey was pre-tested in February 2026 and then fielded through PureSpectrum in March and
April 2026. The final analyzed samples included 1,466 respondents in the United States, 1,494 in Japan, and 1,218 in Sweden.

Results
My study's central finding is that gender quotas increase support for foreign aid to authoritarian regimes across all three donor publics. I show that respondents in the United States, Sweden, and Japan were all more likely to choose countries with parliamentary quotas than those without them. However, the effect of quotas was modest relative to other attributes, such as corruption and women's economic rights, especially in Japan. I also found that quotas increased perceived democracy more than they increased willingness to provide aid, suggesting that respondents do interpret quotas as a democratic signal, but do not automatically translate that signal into support for material assistance.

The most important substantive contribution of the project is that the reputational benefit of quotas is not uniform. Religion strongly conditions how quotas are interpreted. Across all three countries, the positive effect of quotas was systematically weaker when quotas were adopted by Muslim-majority countries, and in some comparisons in Sweden and Japan, the positive effect of quotas was even reversed. By contrast, countries described as having no dominant religion were often evaluated especially favorably. Race, however, showed little evidence of conditioning effects. Overall, the evidence strongly supports the argument that religion shapes how foreign publics interpret democratic signals, while providing much less support for a similar claim about race.

Contributions
My findings contribute to the argument of autocratic genderwashing. My results are consistent with earlier findings in the U.S., while extending support for this argument with new evidence from Sweden and Japan. At the same time, this study connects the literature on gender quotas and autocratic genderwashing to another body of research on religion, race, democracy, and foreign policy opinion. I show that quotas do not "genderwash" equally well across contexts. Their reputational effect depends on who adopts them, and in particular on their religious identity. In this respect, my findings also reinforce prior work demonstrating the importance of
religion in shaping support for democracy and foreign policy preferences. More broadly, my results imply that our understandings of democracy are not completely neutral, but are shaped by underlying identity-based assumptions.

Next Step
I am currently finalizing my thesis manuscript, and once the final version is complete, I will share the finished thesis with the Ethics Institute. I am sincerely grateful for the Institute's support, which made my thesis possible

The Ethics Institute graciously funded my undergraduate honors thesis in Government through the Sayles Research Grant. My thesis examined the central question, "What accounts for variance in how Americans value the lives of foreign civilians relative to how they value in the lives of American soldiers?" Specifically, I looked at the perceived justness of the conflict and the scale of pre-existing civilian fatalities.

To measure this, I ran a survey experiment on Cint and recruited 3,700+ participants to gauge their opinion on this tradeoff question. I asked them to choose between two operations to achieve a strategic military goal in a conflict between the United States and a fictional country called Esor. The operations were identical except that the proportions of soldiers and civilians killed were different. My experiment found that Americans were more likely to choose to save the higher number of civilians at the expense of killing a higher number of American soldiers when the war was considered unjust, operationalized through the United States initiating the conflict unfairly against Esor. I also found that Americans were more likely to choose to save the higher number of civilians at the expense of killing a higher number of American soldiers when there were fewer pre-existing deaths, highlighting how numbing from large numbers affects sensitivity towards civilian casualties through a phenomenon called psychophysical numbing.

My thesis has important implications for American public opinion on foreign policy. It highlights how quickly Americans become desensitized to large numbers of civilian deaths in war and the importance of how elites and the media frame justness when conflicts arise. It also reinforces historical trends analyzed in my case studies examining the Gulf War and the Iraq War, which demonstrate that American acceptance of foreign civilian fatalities remains high, especially when posited against the lives of American soldiers.
The entirety of my funding was used to recruit 3,702 survey participants, giving me significant statistical power to make casual inferences about my hypotheses

1. Project Overview
This research project examined the economic and ethical dimensions of Peru's agricultural transformation. Through nine days of fieldwork across two Peruvian regions, I met with industry executives, visited remote farmlands, spoke with laborers, and even interviewed politicians, including both the former President and Prime Minister of Peru to understand how the Andian nation has navigated ethical tradeoffs in labor conditions and environmental sustainability to become the largest global exporter of table grapes, blueberries, and asparagus.

2. Research Context and Methodology
Peru's agricultural exports have increased more than 30x since 2000, reaching $12.5 billion by 2024. Building on frameworks from Professor Peter DeShazo's GOVT 40.15 (Commodities and Development in Latin America), I designed this project to understand how global market forces shape ethical production decisions. My fieldwork combined site visits to major operations–including IQF del Perú, Agrícola Peru, and Interbank–with interviews spanning the agricultural value chain, from investment bankers to sustainability supervisors to commercial traders and even farmworkers.

In Lima, I met with a plethora of experts in agribusiness, policy, and sustainability. Standout meetings included Andrea Alva Maraví, Agribusiness Director at Interbank, Peru's leading agricultural lender, who explained financing structures enabling sector growth; Dartmouth Trustee Carlos Rodríguez-Pastor T'88 who is CEO of Peru's largest private equity firm; and even former President of Peru Pedro Pablo Kuczynski P'86 who provided a critical perspective on how political stability and governance reforms catalyzed investment flows while stirring labor unrest.

3. Key Findings
3.1 Policy Reforms and Investment
Peru's agricultural transformation began with deliberate policy interventions following decades of instability. Law 27360, passed in 2000, slashed agricultural income taxes from 30% to 15%, subsidized irrigation infrastructure, and relaxed labor regulations. These reforms attracted repatriated Peruvian capital that had been offshored during unstable periods. The scale of investment proved decisive in igniting growth of farming businesses. Unlike Chile's fragmented small-farm model, Peruvian agriculture emerged through institutional capital backing large operations from inception. Farm-establishment costs can reach approximately $100,000 per hectare for table grapes, thus requiring sophisticated financing. This means that while Peruvian farms are some of the most profitable in the world, these fiscal gains accumulate most to an extremely small group of financiers and entrepreneurs. This has led to civil unrest and mass protests, especially as employment protections for farm-workers have remained consistently weak. Critics contend that these mega-farms are responsible for improving working conditions while others argue that these weak-benefits are essential to maintaining jobs and keeping Peru as the world's top exporter.

3.2 Competitive Advantages and Market Displacement

Peru displaced Chile as the leading grape exporter through several interconnected advantages. An interview with John Pandol, a leading grape importer to the U.S. market, emphasized that Peruvian vines reach commercial production in 18 months versus 3-4 years in Chile, enabling rapid adoption of premium varieties from International Fruit Genetics and Sun World International–American GMO companies. Thus, large-scale operations achieve economies-of-scale and profit margins that are impossible for small-farms. At Agrícola, CEO Don Ricardo detailed their vertically integrated model controlling production, packing, and commercialization, enabling rigorous quality control and complete traceability from field to customer. This institutional approach, combined with aggressive recruitment of Chilean agronomists, gave Peruvian growers an unmatched advantage while devastating Chilean farmers as they struggled to keep up.

3.3 Labor Conditions: Employment and Tension
The grape industry alone has generated approximately 100,000 jobs, transforming regions like Ica. During my visit to Agrícola there, I toured a farm with around 11,000 daily workers (of which 45% are female) while they engaged in manual operations of picking and tending to vines. According to Agrícola, they provide transportation, subsidized meals, subsidized housing, and even social programs including schooling. However, there are still apparent gaps in ensuring workers are treated ethically. Recent, 2020 agricultural worker protests resulted in several deaths that exposed tensions over wages and working conditions. The subsequent repeal and revision of Law 27360 reflected precarious balancing between competitiveness and labor protections.

Both Rodríguez-Pastor and Kuczynski emphasized that political stability remains essential for continued investment, yet the government must navigate competing pressures from labor movements, business interests, and environmental advocates. This tension represents an ongoing challenge to sustainable development.

3.4 Environmental Sustainability: The Water Crisis
Water scarcity emerged as the most significant long-term constraint to continued growth. Ica's aquifer, containing roughly 40% of national groundwater reserves, faces overexploitation. Pamela Gómez, sustainability supervisor at Agrícola Cerro Prieto, detailed their transition from wells to surface water via the Gallito Ciego reservoir and canal systems, adopting 100% drip irrigation that reduced consumption by 40%.
Despite technological improvements, aquifer depletion threatens viability. Interbank has responded with sustainability-linked loans with KPIs tied to resource efficiency in an attempt to protect the industry. Nevertheless, a stark tension remains between production growth and water availability.

4. Connection to Coursework
This research applies global value chain analysis from GOVT 40.15 directly to Peru's experience. Peru has achieved process upgrading–efficient large-scale production with modern infrastructure–but not product upgrading (developing proprietary varieties). This positioning maximizes current comparative advantage while limiting value capture and keeping Peru reliant on American companies which dominate the GMO market. From an economics perspective, Peru's success illustrates how policy and infrastructure investment can convert natural endowments into realized competitive advantage. However, it also highlights how companies must sacrifice an entirely ethical approach in order to secure advantage in a global market.

5. Deliverables and Practical Impact
Beyond this report, I am developing a comprehensive PowerPoint presentation that will serve as a strategic deliverable for Peru's agricultural sector leaders. This presentation synthesizes field research, market data, and comparative analysis to help industry leaders and policymakers better understand Peru's economic and environmental positioning within global markets. The presentation will address critical questions about sustainable growth pathways, competitive positioning against Chile and emerging producers, water management imperatives, and strategies for capturing higher value-chain segments through domestic R&D investment and processing capabilities.

6. Conclusion
This research revealed three main ethical tensions in Peru's agricultural transformation. First, weak labor conditions are a tradeoff for export dominance–the grape industry created 100,000 jobs and transformed impoverished regions, yet the 2020 protests resulting in worker deaths exposed real grievances over wages and working conditions. Second, environmental sustainability presents a long-term existential challenge. Currently, even dramatic water efficiency improvements cannot prevent aquifer depletion, revealing that individual corporate virtue may prove insufficient against wide-spread resource constraints. Third, economic dependency on foreign intellectual property and input suppliers limits Peru's autonomy while enabling its competitive advantage, creating strategic vulnerabilities alongside prosperity.
These ethical complexities only became clear through fieldwork. Meeting executives navigating impossible tradeoffs between competitiveness and worker welfare, hearing President Kuczynski explain political calculations behind reform, watching sustainability officers manage water scarcity—these experiences transformed abstract concepts from GOVT 40.15 into urgent human realities. As a Dartmouth economics student, I've studied market failures theoretically. Standing in Ica's fields made them tangible. The presentation I'm developing for Peruvian leaders represents an attempt to bridge academic analysis and practical decision-making. Most importantly, this experience convinced me that meaningful research on development ethics requires engaging directly with communities navigating these tensions daily, not just analyzing from a far-away ivory tower.

Read a summary of Eleanor's research here.

Read Tracy's Thesis here.

Read Samantha's thesis here.

I set about with a set of questions before embarking on this project. Each has served asa guidepost along the routes of this experiential and existential "search and rescue mission."How do I look at historical data? How do I use "creative license?"How does testimony figure scientifically and historically as document? What functions are performed by personal literary narratives and artistic works? What roles can be defined by artists and writers? Who are we to say, and how do we say it? How do we explain interpretation and inspiration as a creative artist? Who is the "authority" and what are the perspectives involved? What position do I reside in, move within, or work beyond? Does this present point in history recall or call for reflection on the past and future?

My project "Sanctuary and Solace" acknowledges the issues of the past and present, conflict and disagreement. This research involves creating bridges for negotiation and understanding, taking form as a portfolio for exhibition of artworks and accompanying written pieces.

Read Nan's reflection essay here.

Read Floating Islands here.

Sample from Nan's portfolio.

Abstract: This study aimed to assess ethical considerations for using digital biomarkers in predicting the onset of neuropsychiatric disease. The study employed semi-structured interviews with 21 people (N = 16 people at-risk for hallucinations and N = 5 current patients who experience visual hallucinations). The interview questions were grounded in the Consolidated Framework for Implementation Research (CFIR). The Rapid and Rigorous Qualitative Data Analysis (RADar) method was used to generate results. Emerging themes included prior knowledge, care partner resources, perception of outcomes, accessibility, interest in the tool, essential safety, organizational capacity, and monetary considerations. Understanding how to address these concerns from a participant's experience may promote engagement and positive attitudes surrounding prodromal biomarker metrics adapted for digital health. Participants mentioned the potential of the digital biomarker to increase access to health care in areas with remote access to in-person health care and offset the trajectory of disease. Participants endorsed coordination with health care providers to provide next steps for modifying the trajectory of disease. Future research should promote team science among researchers, physicians, and individuals with lived experience of mental illness to improve telemedicine practices in accordance with population needs. With the increasing use of telemedicine and other remote health care practices, it is vital to consider the ethics of a stand-alone remote technology to diagnose pre-clinical neuropsychiatric disorders. Read the published paper here.

Family Caregiver Comfort with Telehealth Technologies: Differences by Race and Ethnicity in a Cross-Sectional Survey

Reed Bratches

Background:Telehealth has seen widespread use since the onset of the COVID-19 pandemic, and 82% patients required assistance in accessing their telehealth appointments. This assistance commonly comes from a family caregiver who may or may not be comfortable using the technologies associated with telehealth. The objective of our study was to analyze a demographically representative survey of U.S. family caregivers to understand the level of comfort using telehealth technologies among family caregivers.  Read the published article here.

Exploring the Ethics of Translating Open Biomedical Data into AI-derived Knowledge: From Policy to Practice

Chase Yakaboski,  PhD candidate at Thayer

Chase's research paper will be presented at the NeurIPS AI for Science workshop.

Read the paper:  ArXiv here.

Remembering in Israel

Mira Darham, Graduate Student

Throughout all my courses in the Master of Arts and Liberal Studies Program I have undertaken underlying themes of memory, personal history and the Holocaust. Utilizing the lens of creativity, I make each project I pursue innovative and striking, producing an experience for both the viewer and me. 
 

My goals within this Independent Study are to spend the term in Israel, doing research at Yad Vashem and interviewing survivors, members of the community and my own family. Not long ago my family discovered that we have relatives in France and Israel who survived the Holocaust. This will provide a foundation for my pursuits in personal history and allow the opportunity to actively witness and become reintegrated into my culture and heritage. I will be making frequent visits to Yad Vashem: The World Holocaust Remembrance Center. Predominantly research based, this activity will give historical context and gravity to my pursuits. From its vast collection of archives, testimonies, documents, and databases on the Holocaust, I will glean information on my family and other survivors and victims whose stories I have yet to find. These interviews would complete a trifecta of experience and information-gathering that would address my current studies in creative writing while incorporating themes of cultural studies.  Read Mira's resulting essays here.

Responsible Planning in Assistive Digital Biomarkers: A Qualitative Analysis

Julia Hill '24

Seventy-five million individuals worldwide are impacted due to neuropsychiatric diseases, which include Parkinson's disease, dementia with Lewy bodies, Alzheimer's disease, and schizophrenia spectrum disorders.¹ Their prevalence is expected to double by 2040 as the large baby boomer population ages, with patients dying 10 to 20 years earlier than the general population.² The development and illness trajectory (prognosis) of neuropsychiatric diseases is modifiable if diagnosed early. I feel an urgent need to improve the quality of life and length of life for individuals who suffer from life-altering diseases. Thus, I collaborated with senior researchers: Vinamrita Singh, PhD, Andrew Bohm, PhD, MS, Gajanan Revankar, PhD, MBBS, and Karen L. Fortuna, PhD, LICSW, across multiple disciplines, including sociology, psychology, neuroscience, biology, and physics. I gained an understanding of multidisciplinary research. Then, I produced a novel tool for sustainable change in the disease trajectory. I also mentored two other Dartmouth undergraduates: Morgan Kerber-Folstrom and Margaret Klein, as we constructed the analysis of results. Together, we learned to appreciate the severity of the disease and the importance of scientific methodology in digital biomarker research.  Read more about Julia's research here.

UVelona: A Novel Harm-Reduction Method to Disinfect Needles and Syringes

George Gerber '23, Anahita Tewari Kodali '23, Aksheta Saireddy Kanuganti '24, Joel Smith '24

Injection drug use is a major public health issue in the United States; about 1 million individuals in the US inject drugs every year. Individuals who use injection drugs face a severe lack of resources and extreme stigma. For them, a lack of resources can result in unsafe injection practices, including reuse of needles and syringes. Needle/syringe reuse puts individuals who use injection drugs at a high risk of contracting several blood-borne diseases, including HIV, Hepatitis B, and Hepatitis C. Additionally, because of the strain put on the end of the needle, needle reuse can result in distortion of the needle's fine tip, which can cause the tip to fall off and become embedded in the user's skin. This increases the chance of injury and infection. Read more about their research here.

One Year After the U.S. Capitol Attack: Are Racial and Partisan Identities Opportunistic?

Ella Laurent '25

Affective polarization and racial fluidity are constructs that have become rather prevalent in America's political culture. Based on an online survey experiment with a sample of voting-age adults in the U.S., we examined how the exposure to information about violent protestors with strong racial and partisan influences impacted survey respondents' own racial and partisan identities. Read more about Ella's research here.

The Ethics of Language and Language Use

Zachary Lang '23

My research project was focused on exploring the ethics of language and language use. It did so by following a method that numerous philosophers take when approaching issues in conceptual ethics and conceptual engineering: reverse engineering. Reverse engineering, in essence, consists of assessing various descriptive properties of concepts—for example, their semantic content, their cognitive content, the inferences they generate, and their social functions—and then providing normative and prescriptive assessments.  Read more about Zach's research here.

Ethics in Making "Sustainable Food" in the Palm Oil Supply Chain of a Multilatina Company

Montserrat Perez Castro Perez (Graduate student)

Palm oil is the most widely produced vegetable oil in the world but has received much attention for its implication in the destruction of tropical forests and smallholder livelihoods. The food industry has responded to the negative claims through corporate sustainability [CS] strategies, such as certifications, and initiatives to trace the origin of the raw material, and monitor land-use changes. But implementing CS strategies has not been easy. These are voluntary projects; producers are not contractually obliged, nor are they required by government regulation to provide information. Even with access to data, the control of all the socio-environmental impacts of a value chain is immeasurable. Further, the creation of CS roles and new internal requirements within the organization, are also complicated organizational changes that don't go unchallenged. Due to the growing popularity of CS to approach the socioenvironmental impact of the industry in a voluntary way, is important to understand how ethics inform different actors' views and practices. The research draws from anthropology and social studies of science to ask, what ethical values, practices, and social relationships are involved in making sustainable food out of palm oil?  Read more about Montse's research here.

Mindful User Experiences: The Ethical Future for Persuasive Technologies

Natalie Noelle Svoboda (Graduate student)

Technology is currently designed with very limited ethical standards, leading to the creation of experiences that distract and addict.  This is the result of a carefully-crafted system that is set up for monetary value — the more attention the user dedicates to the experience, the better it is for the business.  As a result, user experience (UX) designers quantify user engagement by measuring screen time, clicks, and shares, keeping users hooked to their screens, treating attention as a commodity.  UX designers create features that ping the user to interact, pressures the user to respond immediately, and provide endless new content for entertainment.  Technology, as a result, is currently manipulating the thoughts and actions of millions of people, resulting in the frazzled and mindless consumption of information. What would our digital world look like if it had an ethical standard to protect user attention?  Read more about Natalie's research here.

Perceptions of the impact of COVID-19 on healthcare communication in a nationally representative cross-sectional survey of family caregivers

The grant supported a nationally-representative survey of family caregivers, and we were able to have a whole section on the COVID-19 pandemic. 

https://bmjopen.bmj.com/content/12/4/e051154

An Introductory Bioethics Podcast for Students at the Geisel School of Medicine and The Dartmouth Institute for Health Policy & Clinical Practice

Ovya Hema Ganesan (Geisel)

There is currently no introductory bioethics podcast available that is targeted to medical and health students. Micro-learning, through the form of a podcast, is becoming more widely used. In particular, podcasts allow students to listen to multiple points of view, follow along with the progression of a thought framework, and avoid reading more content on top of their schoolwork.

We hope to develop a health care ethics podcast meant to increase students' awareness about cases they may face within careers in the healthcare field. Allowing students to develop a framework that can be practiced and applied to real cases is crucial for enabling them to be prepared to face such challenges after graduation.

With each podcast episode, we hope to discuss case-based, practical, ethical issues through an analytical thought process. Through rigorous research and the involvement of expert ethicists, the podcast will cover both novel and classical ethics issues pertinent to health and wellbeing, encouraging listeners to understand that ethical deliberations underlie many of our daily decision-making. Listeners will then be equipped with the framework to address ethical issues in the healthcare field in an equitable fashion. Overall, this podcast seeks to fulfill the aim of the Ethics Institute as an innovative, accessible media: to foster the study and teaching of ethics.

Digital Mental Health Peer Support Intervention for Family Caregivers of People with Dementia

Caroline Collins-Pisano '22 (Thesis)

Family caregivers of people with dementia are critical to the quality of life of care recipients and the sustainability of healthcare systems but face increased risk of emotional distress and negative physical and mental health outcomes. The purpose of this study was to examine the usability, acceptability, and preliminary effectiveness of a technology-based and caregiver-delivered peer support program, "Caregiver Remote Education and Support" (CARES) smartphone/tablet application and to identify barriers and facilitators to the ethical use of former caregivers as technology-based interventionists. Nine family caregivers of people with dementia aged 18 and older received the CARES intervention and three former family caregivers of people with dementia were trained to deliver the CARES system. Quantitative data were collected at baseline and at the end of the two-week field usability study. Qualitative data were also collected at the end of the two-week field usability study. The pilot study demonstrated that a two-week, peer delivered and technology supported mental health intervention designed to improve burden, stress, and strain levels was experienced by former and current family caregivers of people with dementia as usable, acceptable, and ethical. CARES was associated with non-statistically significant improvements in burden, stress, and strain levels. This pre/post field usability study demonstrated it is possible to train former family caregivers of people with dementia to use technology to deliver a mental health intervention to current family caregivers of people with dementia. Future studies would benefit from a longer trial, a larger sample size, a randomized controlled design, and a control of co-variables such as stages of dementia, years providing care, and the severity of dementia symptoms. Read Caroline's thesis here.

Ethical Implications of Media Influence on Altering US Public Opinion

Undergrad student research on how media coverage of the recent demonstrations and violence perpetrated by both the protesters and the pro-China police in Hong Kong affects U.S. public opinion about China and foreign policy preferences towards both China and Hong Kong. Experiment, to be fielded in the U.S., respondents are randomly exposed to either neutral, pro-China, or pro-Hong Kong news about the ongoing demonstrations, and then asked questions about their attitudes and preferences for subsequent U.S. foreign policies.

The State of Social and Political Expression on College Campuses

"What explains variation in student willingness and truthfulness when discussing social and political topics on campus?" Research to determine if a large percentage of undergraduate students across the political spectrum are muting at least some of their beliefs on social or political issues and the reasons behind self-censorship. Some potential reasons include preventing others from being offended, nipping emotionally-exhausting conversations on sensitive issues, being uncomfortable sharing partially-developed thoughts on polarizing topics, or being unwilling to speak on issues in which one lacks personal experience.

Leading Lives and the Normative Significance of Narratives

Xingzhi (Justin) Guo '22

There exists an intuitive connection between living and narrating—when we try to convey to others an understanding of ourselves, we do so by telling the story of our life. An important question ensuing this intuition, then, is whether this connection bears any normative significance upon us—whether narrative concerns matter to our decision-making and actions at all; to what extent can the question of what we should do be reduced to what promotes the narrative coherence of our life story. Some philosophers, like Galen Strawson and Bernard Williams, contended that narratives are without normative importance to us. For the majority of decisions in life, for instance, it is not the case that we consult our life stories up until then and infer what we should do in light of what ought to be done to promote a coherent life story. Rather, actions are often immediate responses to particular circumstances. The relevant considerations concern more immediate projects and ends, and the input/role of large-scale narrative considerations is obscure if not nonexistent. On the other hand, there are philosophers, including Alasdair MacIntyre and Charles Taylor, who argued that we act precisely by pursuing and enacting coherence in the story we are leading. Narrative interests would therefore be decisive in guiding our actions. Indeed, when it comes to important decisions in life, such as career, marriage, and parenthood, the question in the end is always how the decision would impact and/or shed light on the story that our life would amount to (whether, for example, as the result of our choices, our life would become a story of redemption and improvement). 

How can we adjudicate between these two positions? Both intuitions seem real and important, and a satisfying account would need to incorporate both. In my project, I want to evaluate two proposals (one coming from each position) that have great promises in addressing this concern. Holding on to Strawson and Williams's position, the first proposal argues that the fact that we tend to refer back to our life stories at certain moments of decision needs not imply that life narrative is somehow indispensable to making decisions and taking actions. Rather, this may be explained in terms of social conventions: we are prescribed the norm of taking narrative considerations into view when it comes to certain moments of decisions, but it is still an open question whether it is inconceivable for us to decide and act in those moments without taking an interest in our life narratives. Another proposal, defending MacIntyre and Taylor's position, argues that the fact that we do not appeal to narrative concerns in each and every action is not sufficient to show that life narratives are irrelevant to ordinary decision making. Indeed, narrative considerations may play a reactive role insofar as it sets the boundaries for actions. It may delineate possible actions that would be considered out of bound in light of their disruption of our narrative coherence. Akin to laws or commandments, the demands of narrative coherence may not legislate positive content for what we ought to do from moment to moment. Nevertheless, it supervises and shapes all actions as it determines what ought not to be done. That narrative considerations do not surface in day-to-day actions would only mean that ordinarily our intentional activities do not threaten our narrative coherence.

I seek to take an interdisciplinary approach to evaluating the two proposals. Because this is mainly a question in contemporary philosophy of action and ethics, I will further delve into important discussions raised by David Velleman (in his book Self to Self as well as in his articles Narrative Explanations, The Self as Narrator, among others). I will also consider relevant chapters in Alasdair MacIntyre's After Virtue and Charles Taylor's Sources of the Self. Beyond (just) philosophy, I want to explore the intersection between philosophy and literary theory, where much theoretical work on narratives in literature has been done. This would encompass Paul Ricoeur's Time and Narrative, Walter Benjamin's Illuminations, and Aristotle's Poetics. Lastly, incorporating the empirical perspective, I seek to include readings in contemporary developmental psychology (works on autobiographical selfhood by Qi Wang) and conduct case studies on the nature of life stories (drawing from the life story database of NPR'S Story Corps). I intend to conduct all of the above through self-initiated reading clubs so that reading and thinking about narratives will not be an impersonal research project but a personal, reflective, and interactive experience.