arXiv subjects

Sophia Kazinnik

Publications and source records attributed to Sophia Kazinnik.

4 recordsLinked to original sources

arXiv2026

AISPA: User-Centric System Prompt Auditing for Large Language Model Applications

System prompts are instructions configured by developers to govern the behaviors of foundation models in AI applications. They are used throughout commercial AI products, but are rarely disclosed to the public or regulators, creating a serious trust and accountability gap in the wide deployment of AI systems. In this paper, we introduce Artificial Intelligence System Prompt Assurance (AISPA), a user-centric framework for systematically auditing system prompts in AI systems. AISPA examines specific parts of a system prompt and evaluates them along eight dimensions that matter to users. We then use this framework to review 3,249 instructions from system prompts in 88 commercial AI products, classifying each instruction as either protective (of users) or problematic. Our audit surfaces four core findings. First, system prompt design varies substantially across products and developers, with some organizations averaging over 60 protective instructions per product while others average fewer than 5. Second, protective instructions are widely adopted but shallow in scope: 98.9% of products contain at least one, yet only 24% cover all eight dimensions of the AISPA taxonomy. Third, system prompts have grown steadily longer and more protective of users, suggesting that user protection is becoming a more visible concern in commercial prompt design. Fourth, despite this progress, problematic instructions remain pervasive: roughly 40% of products contain at least one instruction that works against user interests, and protective and problematic instructions frequently coexist within the same prompt. Our findings highlight the need for greater transparency, standardization, and independent oversight for system prompts in commercial AI products.

cs.AI↗

arXiv2025

Federal Reserve Communication and the COVID-19 Pandemic

In this study, we examine the Federal Reserve's communication strategies during the COVID-19 pandemic, comparing them with communication during previous periods of economic stress. Using specialized dictionaries tailored to COVID-19, unconventional monetary policy (UMP), and financial stability, combined with sentiment analysis and topic modeling techniques, we identify a distinct focus in Fed communication during the pandemic on financial stability, market volatility, social welfare, and UMP, characterized by notable contextual uncertainty. Through comparative analysis, we juxtapose the Fed's communication during the COVID-19 crisis with its responses during the dot-com and global financial crises, examining content, sentiment, and timing dimensions. Our findings reveal that Fed communication and policy actions were more reactive to the COVID-19 crisis than to previous crises. Additionally, declining sentiment related to financial stability in interest rate announcements and minutes anticipated subsequent accommodative monetary policy decisions. We further document that communicating about UMP has become the "new normal" for the Fed's Federal Open Market Committee meeting minutes and Chairman's speeches since the Global Financial Crisis, reflecting an institutional adaptation in communication strategy following periods of economic distress. These findings contribute to our understanding of how central bank communication evolves during crises and how communication strategies adapt to exceptional economic circumstances.

econ.GN↗

arXiv2025

Social Group Bias in AI Finance

Financial institutions increasingly rely on large language models (LLMs) for high-stakes decision-making. However, these models risk perpetuating harmful biases if deployed without careful oversight. This paper investigates racial bias in LLMs specifically through the lens of credit decision-making tasks, operating on the premise that biases identified here are indicative of broader concerns across financial applications. We introduce a reproducible, counterfactual testing framework that evaluates how models respond to simulated mortgage applicants identical in all attributes except race. Our results reveal significant race-based discrepancies, exceeding historically observed bias levels. Leveraging layer-wise analysis, we track the propagation of sensitive attributes through internal model representations. Building on this, we deploy a control-vector intervention that effectively reduces racial disparities by up to 70% (33% on average) without impairing overall model performance. Our approach provides a transparent and practical toolkit for the identification and mitigation of bias in financial LLM deployments.

econ.GN↗

arXiv2025

Measuring Communication Quality of Interest Rate Announcements

We use text-mining techniques to measure the accessibility and quality of information within the texts of interest rate announcements published by the Bank of Israel over the past decade. We find that comprehension of interest rate announcements published by the Bank of Israel requires fewer years of education than interest rate announcements published by the Federal Reserve and the European Central Bank. In addition, we show that the sentiment within these announcements is aligned with economic fluctuations. We also find that textual uncertainty is correlated with the volatility of the domestic financial market.

econ.GN↗