RePEcML & AI Methods

Do AI Assistants Reinforce Investor Beliefs? Evidence from Human and AI Replies

AI-generated investment replies are systematically more bullish than human replies on Reddit; instruction-tuning amplifies bullishness beyond the base model's inherent bias.

Read the paper PDF

View as Markdown Open in Claude Open in ChatGPT Open in PerplexitySite index for AI (llms.txt)

Featured in No. 137 on 30 Oct 2026 · 4 days after release

Figure 1: AI-human bullish gap: instruction-tuned model versus human replies. Panel A plots mean pA(Post+Reply) against pA(Post) in 25 equal-width bins; Panel B plots the gap ∆= pA(Post+AI) −¯pA(Post+Human). Shaded bands are ±1 SE. Instruct probe; N = 15,206 posts.

Released: 26 Oct 2026
First featured: No. 137 · 30 Oct 2026
Published in: Not yet, as far as Semantic Scholar knows
Fanfare: 2 of 5
Identifier: RePEc:bfi:wpaper:2026-133
Authors: Swaminathan Balasubramaniam et al.

Citations and venue from Semantic Scholar (ODC-BY), refreshed weekly. Summary: Quant Letter (CC BY 4.0).

Related papers

Adjoint Matching: Fine-tuning Flow and Diffusion Generative Models with Memoryless Stochastic Optimal Control

Scaling Proprioceptive-Visual Learning with Heterogeneous Pre-trained Transformers

Continual Diffusion: Continual Customization of Text-to-Image Diffusion with C-LoRA

Fine-tuning can cripple your foundation model; preserving features may be the solution

LLARVA: Vision-Action Instruction Tuning Enhances Robot Learning

DITTO: Diffusion Inference-Time T-Optimization for Music Generation