The Indian Financial Automated Advice Bias (IFAAB) Dataset : Auditing Socioeconomic and Demographic Disparities in LLM-Generated Financial Advice in India

Kriti Mahajan

cs.CYOct 28, 2026

The Indian Financial Automated Advice Bias (IFAAB) Dataset : Auditing Socioeconomic and Demographic Disparities in LLM-Generated Financial Advice in India

Authors: Kriti Mahajan

Organizations: This work was conducted independently in the author’s personal capacity and was self-funded, without financial support from Amazon. The views and conclusions expressed herein are solely those of the author and do not represent those of Amazon or its affiliates.

Abstract

Large language models are increasingly used for financial guidance, yet little is known about whether their advice changes across social identities in Global South contexts. We introduce the Indian Financial Automated Advice Bias (IFAAB) benchmark, a counterfactual audit dataset of 31,734 GPT-4o-generated financial-advice responses for synthetic Indian personas varying by gender, caste, education, employment, residence, age, and district. Holding the underlying request fixed, we evaluate disparities in LLM-implied wealth, income, savings, projected gains, textual similarity, advice topics, and qualitative bias judgments. In the gender-only audit, male-labelled profiles receive 60.1% higher initial total resources and 46.8% higher post-advice total resources than comparable female-labelled profiles; in the caste-augmented audit, the post-advice total-resource gap remains 49.4%. Adding district information reduces the aggregate post-advice total-resource gender gap to 7.9%, revealing strong geographic conditioning. Standalone caste comparisons also show systematic stratification: relative to General-category profiles, OBC, SC, and ST profiles receive significantly lower post-advice total resources, respectively. Textual audits show low semantic similarity between paired gendered responses, with an average TF--IDF cosine similarity score of 0.44. Qualitative analysis shows that women are more often steered toward cautious saving, community support, and basic financial literacy, while men receive more growth-oriented advice involving investment, entrepreneurship, and asset accumulation. IFAAB provides a reusable benchmark for evaluating fairness, suitability, and consumer-protection risks in LLM-mediated financial advice, and highlights the need for locally grounded, intersectional evaluation of AI systems in high-stakes economic domains.

Figures & tables

Panel of Figure 1 — Figure 1: Distribution of gains by gender across wealth, income, savings, and total resources.

Figure 2: Mean total gain differences by education, employment, caste, and location.

Figure 3: District-level male–female percentage differences in initial and post-advice financial outcomes.Values report male-female differences; percentages are calculated as (male mean - female mean) / female mean

Appendix figures & tables11 assets

Supplementary material from the paper’s appendix.

Appendix

Figure 4: District-level population financial outcomes for initial and post-advice estimates.

Figure 5: Mean income gain differences by education, employment, caste, and location.

Figure 6: Mean savings gain differences by education, employment, caste, and location.

Figure 7: Mean wealth gain differences by education, employment, caste, and location.

Figure 8: LLM-as-judge topic bias assessment across financial advice responses.

Demographic	Category value	N	Bias rate (%)
Employment	Unspecified	234	96.2%
Employment	Employed	234	97.0%
Employment	Unemployed	234	96.2%
Education	PhD	117	94.0%
Education	Unspecified	117	96.6%
Education	College Educated	117	95.7%

Appendix

Table 15: IFAAB Gender: LLM-as-judge bias-with-difference rates by demographic group.

Employment	Education	Location	N	Bias rate (%)
Unspecified	Phd	Unspecified	13	92.3%
Unspecified	Phd	Rural	13	100.0%
Unspecified	Phd	Urban	13	92.3%
Unspecified	Unspecified	Unspecified	13	84.6%
Unspecified	Unspecified	Rural	13	84.6%
Unspecified	Unspecified	Urban	13	100.0%

Appendix

Table 16: IFAAB Gender: LLM-As-A-Judge Bias Rates by Employment, Education, and Location

Demographic	Category value	N	Bias rate (%)
Employment	Unspecified	3886	98.9%
Employment	Employed	3887	93.2%
Employment	Unemployed	3889	98.0%
Education	Unspecified	2330	99.4%
Education	College Educated	2331	96.4%
Education	High School Educated	2333	99.3%

Appendix

Table 17: IFAAB Gender x District: LLM-as-judge bias-with-difference rates by demographic group.

Employment	Education	N	Bias rate (%)
Unspecified	Unspecified	777	100.0%
Unspecified	College Educated	777	97.9%
Unspecified	High School Educated	777	99.0%
Unspecified	Post-Graduate	777	97.9%
Unspecified	Uneducated	778	99.9%
Employed	Unspecified	777	99.6%

Appendix

Table 18: IFAAB Gender x District: LLM-As-A-Judge Bias Rates by Employment and Education

Advice category	Observed bias pattern	Example 1	Example 2
Banking and Inclusion	Women are more often directed toward local banks, post offices, or basic-access products, while men are given more standard banking or return-oriented options.	Male: “Open a savings account with a bank that offers a good interest rate and no minimum balance requirement.” Female: “Open a savings account with a local bank or post office to keep your savings secure.”	Male: “Open a recurring deposit account to instill a habit of saving regularly.” Female: “Use a high-interest savings account or a liquid mutual fund to park emergency funds for easy access.”
Budgeting and Expense Management	Men are often given more numerical or target-based budgeting rules, while women receive more conservative expense-control advice.	Male: “Aim to reduce discretionary spending by at least 20%.” Female: “Reduce discretionary spending and focus on essential expenses.”	Male: “Allocate 50% of your income to essentials, 20% to savings, and 30% to discretionary spending.” Female: “Aim to limit non-essential expenses to 20% of your income.”
Community Engagement	Men are more often advised to network for jobs or apprenticeships, while women are directed toward self-help groups, cooperatives, or local support networks.	Male: “Network within the community to find apprenticeship opportunities or entry-level jobs.” Female: “Join local SHGs to access microloans and savings opportunities.”	Male: “Participate in local community programs that may offer stipends or small earnings.” Female: “Engage with local NGOs and self-help groups for financial literacy and income-generating activities.”
Community Support	Women receive more gender-specific community-support advice, while men receive more general employment or networking support.	Male: “Engage with local community groups or NGOs for financial assistance or part-time work opportunities.” Female: “Reach out to local NGOs for support programs for elderly women in rural areas.”	Male: “Engage with local community centers or NGOs that offer job placement services.” Female: “Investigate local government programs for financial aid or employment support tailored for women.”
Earning and Income Opportunities	Men are directed toward formal, external, or sector-based employment, while women are often directed toward home-based, informal, or traditionally feminized work.	Male: “Secure a job in retail, hospitality, or any sector with high demand for entry-level positions.” Female: “Consider vocational training in tailoring, beauty services, or handicrafts.”	Male: “Consider part-time or freelance work to gain experience and income.” Female: “Explore part-time or freelance opportunities that can be done from home, such as data entry or customer service roles.”
Education	Men are more often encouraged toward advanced or career-enhancing education, while women are sometimes framed as needing basic literacy, numeracy, or limited skill training.	Male: “Focus on skills that are in demand, such as digital marketing, coding, or data analysis.” Female: “Focus on basic literacy, numeracy, and digital skills.”	Male: “Consider upskilling through online courses or certifications relevant to your job.” Female: “Consider upskilling through online courses in digital marketing, data entry, or basic computer skills.”

Appendix

Table 20: Examples of Gendered Differences in Financial Advice by Advice Category

Panel of Figure 9 — Figure 9: Distribution of Initial and Post-Advice LLM Generated Financial Variables

Demographic Bias in Language Models

Demographic bias in language models studies how identity cues shape model outputs, decisions, and representations, and how to detect and reduce resulting disparities.

See all 137 papers

RupeeBias: Auditing Demographic Bias in Indian Economic Guidance from Large Language Models
Sep 25, 2026Pavithra P M Nair, Bhavik Talaviya, Shourya Bhushan +5Language Model Bias Evaluation Demographic Bias in Language Models
Center for Cybersecurity Systems & Networks, Amrita Vishwa Vidyapeetham, Amritapuri, India · Unique Identification Authority of India, Delhi, India
Thin Evidence, Thick Priors: How Language Models Substitute Identity for Missing Financial Facts
Oct 6, 2026Saanvi Khetan, Sankar BalasubramanianDemographic Bias in Language Models LLM Auditing
The International School, Bangalore · Indian Institute of Science (IISc), Bangalore
Scoring-Rule Sensitivity in LLM Evaluation: Evidence from Indian Financial Regulatory Text
Apr 21, 2026Rajveer Singh PallLLM Evaluation Financial QA
Gyan Ganga Institute of Technology and Sciences, Jabalpur, India

Sep 25, 2026J/K move · Enter open · S save

RupeeBias: Auditing Demographic Bias in Indian Economic Guidance from Large Language Models

Pavithra P M Nair, Bhavik Talaviya, Shourya Bhushan, Rahul Pankajakshan+4

Center for Cybersecurity Systems & Networks, Amrita Vishwa Vidyapeetham, Amritapuri, India · Unique Identification Authority of India, Delhi, India

Language Model Bias Evaluation Demographic Bias in Language Models

Open paper AskPDF arXiv

Individuals turn to large language models (LLMs) for guidance across a wide range of economic tasks, from comparing loan options and planning savings to deciding what raise to ask for or how much to charge for their services. LLMs are known to reproduce social biases, and biased economic guidance may influence what users believe they are worth, what they ask for, and what they ultimately accept. This risk is especially salient in India, where economic outcomes are shaped by demographic categories such as caste and urban-rural location. Existing LLM bias benchmarks, however, are largely designed around Western demographic categories and therefore miss key axes of economic disparity in the Indian context. We introduce RupeeBias, a benchmark for auditing demographic bias in LLM-generated economic guidance across Indian economic settings. RupeeBias consists of 39,150 prompts spanning four use cases: salary estimation, salary increment estimation, counter-offer recommendation, and service pricing recommendation. The benchmark follows a single-attribute counterfactual design, holding the description of the user's qualifications, experience, or service offering fixed while varying one demographic identifier at a time. RupeeBias covers 87 India-specific demographic identifiers across six axes: caste, religion, regional identity, gender, disability, and urban-rural location, with all prompts constructed in both English and Hinglish. We evaluate nine LLMs on RupeeBias and find systematic demographic disparities across all six axes. For otherwise identical prompts that differ only in demographic identifier, LLM-generated economic outputs differ by 20.2% on average. We publicly release RupeeBias to support future research on demographic bias in LLM-generated economic guidance across India-specific demographic and economic contexts.

Abstract

Demographic Bias in Language Models

Explore similar work

RupeeBias: Auditing Demographic Bias in Indian Economic Guidance from Large Language Models

Thin Evidence, Thick Priors: How Language Models Substitute Identity for Missing Financial Facts

Scoring-Rule Sensitivity in LLM Evaluation: Evidence from Indian Financial Regulatory Text