This paper examines the potential of large language models to mimic human survey respondents and derive their preferences.

Massive open online courses (MOOCs) have the potential to democratize education by improving access. Although retention and completion rates for nonpaying users not been promising, these statistics are much brighter who pay receive a certificate upon completing course. We investigate whether paying option can increase engagement with course content. In particular, we consider two effects: (1) effect, which is boost in motivation stay engaged certificate; (2) sunk-cost arises solely because...

Using a field experiment on Pandora Media, the researchers show that one can increase subscription profits by 7% personalizing ad load across users without decreasing profits.

A bias-correction approach that can recover the total average treatment effect of a ranking algorithm based on past A/B tests even if those suffer from interference issues.

The role of advertising as an ``implicit price'' has long been recognized by economists and marketers. However, the impact personalizing implicit prices on firm profits consumer welfare not studied. We first conduct a set large-scale field experiments Pandora exogenously shifting ad load, that is price, for ad-supported users. then use state-of-the-art machine-learning model to examine heterogeneous treatment effects firm's interventions listeners behavior, both in terms listening hours...

Language has a strong influence on our perceptions of time and rewards. This raises the question whether large language models, when asked in different languages, show preferences for rewards over if their choices are similar to those humans. In this study, we analyze responses GPT-3.5 (hereafter referred as GPT) prompts multiple exploring between smaller, sooner larger, later Our results that GPT displays greater patience prompted languages with weak future tense references (FTR), such...

We evaluate whether and how branded TV product placement affects sales for cigarette brands.

This study documents the impact of different forms advertising on demand for cigarettes and their substitutes. We examine effects educational public service announcements (PSAs), e-cigarette advertising, direct-to-consumer (DTC) tobacco cessation prescription drugs, by nicotine replacement therapies (NRTs) such as patches gums cigarettes, e-cigarettes, NRTs, smoking drugs. Among these we find evidence only effectiveness DTC drugs in reducing cigarette demand. increased new prescriptions...

Bans on menthol cigarettes are ineffective if consumers can bypass them. Taxes might be better because they reduce smoking and generate more tax revenue.

Current practice for treatment planning optimization can be both inefficient and time consuming. In this paper, we propose an automated methodology that aims to combine explorative prescriptive approaches improving the efficiency quality of process. Given a plan, our approach explores trade-offs between different objectives finds acceptable region objective function weights via inverse optimization. Intuitively, shape size these regions describe how 'sensitive' patient is perturbations in...

Public health regulators in the United States are currently advocating for a ban on menthol-flavored cigarettes because they believed to be more dangerous than traditional non-menthol cigarettes. However, these bans will have limited benefits if consumers able circumvent them. We examine this issue by evaluating effects of statewide menthol that was instituted Massachusetts 2020. An examination store-level retail sales data from indicates some demand shifted menthols non-menthols after...

The paper measures the cross-category spillover effects of a retailer changing its assortment at extensive margin (by dropping an entire category from portfolio) on outcomes for rivals in industry.

Content platforms routinely face the important challenge of managing user attrition. Usually, attrition involves investing in content improvement, altering monetization approach, or employing promotions. In this paper, we investigate whether can leverage engagement with discussion forums as a vehicle for retaining customers. We use unique dataset from Coursera, leading online education platform, study. Coursera's core offering consists course material that includes lecture videos and...

A randomized experiment with almost 35 million Pandora listeners enables us to measure the sensitivity of consumers advertising, an important topic study in era ad-supported digital content provision. The into nine treatment groups, each which received a different level audio advertising interrupting their music listening, highest group receiving more than twice as many ads lowest group. By maintaining consistent assignment for 21 months, we long-run demand effects and find ad-load three...

We measure the cross-category spillover effects of a retailer changing its assortment at extensive margin (by dropping an entire category from portfolio) on outcomes for rivals in industry.By leveraging quasi-experimental nature exit by national pharmacy chain, hereinafter referred to as Exiting Chain (EC) tobacco category, we this event revenue generated non-tobacco products rival chains. show, using Nielsen RMS data, that each 1% increase cigarettes sales non-EC stores are located near...

Online marketplaces use ranking algorithms to determine the rank-ordering of items sold on their websites. The standard practice is optimal algorithm using A/B tests. We present a theoretical framework characterize Total Average Treatment Effect (TATE) in an test and show that naive TATE estimates can be biased due interference. propose bias-correction approach recover based past tests, even if those tests suffer from combination interference issues. Our solution leverages data across...

Purpose: To develop an automated planning methodology that exploits patient sensitivity to objective function weights. Methods: Given a treatment plan, we first create acceptable region encompasses set of plans with similar clinical performance (e.g., +/−1% at V70Gy). We use inverse optimization map this in criterion space the weight and find corresponding vectors (W). The shape size W describes how sensitive is perturbations exploit information encoded by these regions, approximate for each...

This research analyzes on-screen gender diversity among cable news networks in the United States, focusing on three major (CNN, Fox News, and MSNBC) from 2010 to 2021. We examine amount of time women are shown screen compared men, topics asked discuss their male counterparts, effect talent viewership. also investigate how these patterns differ a number important dimensions: across networks, through time, between different types programs, vs. female viewers, depending viewers' familiarity...

Massive Open Online Courses (MOOCs) have the potential to democratize education by improving access. Although retention and completion rates for non-paying users not been promising, these statistics are much brighter who pay receive a certificate upon completing course. We investigate whether paying option can increase engagement with course content. In particular, we consider two such effects: (a) effect, which is boost in motivation stay engaged order certificate; (b) sunk-cost arises...