Publications
Publications
Show Results For
- All HBS Web
(3,486)
- Faculty Publications (854)
Methods
→
- 2023
- Article
MoPe: Model Perturbation-based Privacy Attacks on Language Models
By: Marvin Li, Jason Wang, Jeffrey Wang and Seth Neel
Recent work has shown that Large Language Models (LLMs) can unintentionally leak sensitive information present in their training data. In this paper, we present Model Perturbations (MoPe), a new method to identify with high confidence if a given text is in the training... View Details
- 2023
- Article
Post Hoc Explanations of Language Models Can Improve Language Models
By: Satyapriya Krishna, Jiaqi Ma, Dylan Slack, Asma Ghandeharioun, Sameer Singh and Himabindu Lakkaraju
Large Language Models (LLMs) have demonstrated remarkable capabilities in performing complex tasks. Moreover, recent research has shown that incorporating human-annotated rationales (e.g., Chain-of-Thought prompting) during in-context learning can significantly enhance... View Details
- 2023
- Other Article
The Harvard USPTO Patent Dataset: A Large-Scale, Well-Structured, and Multi-Purpose Corpus of Patent Applications
By: Mirac Suzgun, Luke Melas-Kyriazi, Suproteem K. Sarkar, Scott Duke Kominers and Stuart Shieber
Innovation is a major driver of economic and social development, and information about many kinds of innovation is embedded in semi-structured data from patents and patent applications. Though the impact and novelty of innovations expressed in patent data are difficult... View Details
- 2023
- Article
Verifiable Feature Attributions: A Bridge between Post Hoc Explainability and Inherent Interpretability
By: Usha Bhalla, Suraj Srinivas and Himabindu Lakkaraju
With the increased deployment of machine learning models in various real-world applications, researchers and practitioners alike have emphasized the need for explanations of model behaviour. To this end, two broad strategies have been outlined in prior literature to... View Details
- 2023
- Article
Which Models Have Perceptually-Aligned Gradients? An Explanation via Off-Manifold Robustness
By: Suraj Srinivas, Sebastian Bordt and Himabindu Lakkaraju
One of the remarkable properties of robust computer vision models is that their input-gradients are often aligned with human perception, referred to in the literature as perceptually-aligned gradients (PAGs). Despite only being trained for classification, PAGs cause... View Details
- November–December 2023
- Article
Iterative Coordination and Innovation: Prioritizing Value over Novelty
By: Sourobh Ghosh and Andy Wu
An innovating organization faces the challenge of how to prioritize distinct goals of novelty and value, both of which underlie innovation. Popular practitioner frameworks like Agile management suggest that organizations can adopt an iterative approach of frequent... View Details
- 2023
- Working Paper
Causal Interpretation of Structural IV Estimands
By: Isaiah Andrews, Nano Barahona, Matthew Gentzkow, Ashesh Rambachan and Jesse M. Shapiro
We study the causal interpretation of instrumental variables (IV) estimands of nonlinear, multivariate structural models with respect to rich forms of model misspecification. We focus on guaranteeing that the researcher's estimator is sharp zero consistent, meaning... View Details
Keywords: Mathematical Methods
- October 2023
- Article
Improving Regulatory Effectiveness Through Better Targeting: Evidence from OSHA
By: Matthew S. Johnson, David I. Levine and Michael W. Toffel
We study how a regulator can best target inspections. Our case study is a U.S. Occupational Safety and Health Administration (OSHA) program that randomly allocated some inspections. On average, each inspection averted 2.4 serious injuries (9%) over the next five years.... View Details
- 2023
- Working Paper
In-Context Unlearning: Language Models as Few Shot Unlearners
By: Martin Pawelczyk, Seth Neel and Himabindu Lakkaraju
Machine unlearning, the study of efficiently removing the impact of specific training points on the
trained model, has garnered increased attention of late, driven by the need to comply with privacy
regulations like the Right to be Forgotten. Although unlearning is... View Details
- September 2023
- Module Note
Live Case Exercise for Financial Reporting
By: Tatiana Sandino and Marshal Herrmann
Harvard Business School employs the case method as a cornerstone of its pedagogy, providing students with opportunities to engage in discussions related to difficult or contentious decisions confronted by real-world organizations. In this “live case,” we depart from... View Details
- September 2023
- Technical Note
Note on Difficult Conversations in the Family Enterprise
The best time to have a difficult conversation is, ideally, as soon as possible. Engaging in challenging conversations early can produce beneficial results for several reasons, such as resolving issues, improving communication, preserving relationships, and increasing... View Details
- September 2023 (Revised January 2024)
- Case
Helmy Abouleish: Making a Desert Bloom
By: Geoffrey G. Jones and Maxim Pike Harrell
This case examines the history of prominent Egyptian-based social enterprise SEKEM from its foundation in 1977 until the COP27 conference held in Sharm El-Sheikh in 2022. Led by father and son team Ibrahim and Helmy Abouleish, SEKEM turned desert into farmland using... View Details
- September–October 2023
- Article
Interpretable Matrix Completion: A Discrete Optimization Approach
By: Dimitris Bertsimas and Michael Lingzhi Li
We consider the problem of matrix completion on an n × m matrix. We introduce the problem of interpretable matrix completion that aims to provide meaningful insights for the low-rank matrix using side information. We show that the problem can be... View Details
Keywords: Mathematical Methods
- September 2023
- Article
Measuring Time Use in Rural India: Design and Validation of a Low-Cost Survey Module
By: Erica Field, Rohini Pande, Natalia Rigol, Simone Schaner, Elena Stacy and Charity Troyer Moore
Time use data can help us understand individual labor supply choices, especially
for women who often provide unpaid care and home production. Although
enumerator-assisted diary-based time use data collection is suitable for
low-literacy populations, it is costly and... View Details
- 2023
- Article
On Minimizing the Impact of Dataset Shifts on Actionable Explanations
By: Anna P. Meyer, Dan Ley, Suraj Srinivas and Himabindu Lakkaraju
The Right to Explanation is an important regulatory principle that allows individuals to request actionable explanations for algorithmic decisions. However, several technical challenges arise when providing such actionable explanations in practice. For instance, models... View Details
- October 2023
- Article
What Does the Inflation Reduction Act Mean for Patients and Physicians?
By: Amitabh Chandra and Benedic Ippolito
The debate around prescription drug measures in the recently passed U.S. Inflation Reduction Act (IRA), which limit some patients’ out-of-pocket costs, has not fully addressed their effect on physicians and patients via their effect on payers. Reducing patients’ costs... View Details
- July–August 2023
- Article
Demand Learning and Pricing for Varying Assortments
By: Kris Ferreira and Emily Mower
Problem Definition: We consider the problem of demand learning and pricing for retailers who offer assortments of substitutable products that change frequently, e.g., due to limited inventory, perishable or time-sensitive products, or the retailer’s desire to... View Details
- August 2023
- Article
Explaining Machine Learning Models with Interactive Natural Language Conversations Using TalkToModel
By: Dylan Slack, Satyapriya Krishna, Himabindu Lakkaraju and Sameer Singh
Practitioners increasingly use machine learning (ML) models, yet models have become more complex and harder to understand. To understand complex models, researchers have proposed techniques to explain model predictions. However, practitioners struggle to use... View Details
- 2023
- Working Paper
How People Use Statistics
By: Pedro Bordalo, John J. Conlon, Nicola Gennaioli, Spencer Yongwook Kwon and Andrei Shleifer
We document two new facts about the distributions of answers in famous statistical problems: they are i) multi-modal and ii) unstable with respect to irrelevant changes in the problem. We offer a model in which, when solving a problem, people represent each hypothesis... View Details
- 2023
- Article
Towards Bridging the Gaps between the Right to Explanation and the Right to Be Forgotten
By: Himabindu Lakkaraju, Satyapriya Krishna and Jiaqi Ma
The Right to Explanation and the Right to be Forgotten are two important principles outlined to regulate algorithmic decision making and data usage in real-world applications. While the right to explanation allows individuals to request an actionable explanation for an... View Details