Hiroshi Takahashi
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Hiroshi Takahashi has a total of 70 co-authors. Here are the top co-authors by number of shared publications:
- Takamasa Kikuchi
Chiba University of Commerce
7 shared publications - Eiji Murakami
Keio University
4 shared publications - Masaaki Kunigami
Tokyo Institute of Technology
4 shared publications - Takao Terano
CUC Research Institute, Chiba University of Commerce, Chiba, Japan
4 shared publications - Eiji Murakami
Keio University
3 shared publications - Sejun Woo
Keio University
3 shared publications
Latest Publications
- 6 October 2026
- Published by Springer Nature
- p. 73-84
Building Personas and LLM Assessment Indicators: For Frailty Prevention
- 6 October 2026
- Published by Springer Nature
- p. 73-84
Building Personas and LLM Assessment Indicators: For Frailty Prevention
- 6 October 2026
- Published by Springer Nature
- p. 97-108
Quantitative Persona Generation Through Large Language Model and Hierarchical Classification for Simulation Results
- 6 October 2026
- Published by Springer Nature
- p. 97-108
Quantitative Persona Generation Through Large Language Model and Hierarchical Classification for Simulation Results
- 6 October 2026
- Published by Springer Nature
- p. 109-120
Improving LPG Supply Network Recovery Through Data-Driven BCP: Quantitative Evaluation of Regional Characteristics in the Philippines
- 6 October 2026
- Published by Springer Nature
- p. 109-120
Improving LPG Supply Network Recovery Through Data-Driven BCP: Quantitative Evaluation of Regional Characteristics in the Philippines
- 2 September 2026
- Published by Emerald Publishing in Journal of Property Investment & Finance
- p. 1-15
Impact of appraisers' façade sensory evaluations on office property valuation: evidence from structural analyses of J-REIT office buildings
Purpose: This study investigates how real estate appraisers' sensory evaluations of office building façades influence property valuation. By identifying the latent factor structure underlying façade impressions and examining both their direct and indirect effects on appraisal values, this study empirically reveals appraisal professionals' tacit knowledge that has traditionally remained unquantified in property valuation practice. Design/methodology/approach: The analysis focuses on office buildings in Tokyo held by Japanese Real Estate Investment Trusts (J-REITs) and appraised by professionals from Japan's largest real estate appraisal organisation. Multiple regression analysis is employed to compare property valuation models with and without façade sensory evaluation variables, while path analysis is used to identify the indirect effects of façade-related components, such as building colour, on property value. Findings: Façade sensory evaluation significantly improves the explanatory power of property valuation models and influences property values at the 1% significance level. Among building attributes, the number of floors has the strongest effect on façade evaluation, followed by colour saturation, building age and brightness. Meanwhile, building colour indirectly affects property value through appraisers' façade evaluations. Practical implications: By quantifying appraisers' tacit knowledge of façade impressions, this study clarifies sources of information asymmetry in property markets and highlights the importance of perceived building scale and colour in professional office property valuation. Originality/value: Through path analysis that incorporates building colour, this study uniquely demonstrates how real estate appraisers' façade impressions structurally influence property value. In doing so, it extends existing research on façade design and property valuation.
- 2 September 2026
- Published by Emerald Publishing in Journal of Property Investment & Finance
- p. 1-15
Impact of appraisers' façade sensory evaluations on office property valuation: evidence from structural analyses of J-REIT office buildings
Purpose: This study investigates how real estate appraisers' sensory evaluations of office building façades influence property valuation. By identifying the latent factor structure underlying façade impressions and examining both their direct and indirect effects on appraisal values, this study empirically reveals appraisal professionals' tacit knowledge that has traditionally remained unquantified in property valuation practice. Design/methodology/approach: The analysis focuses on office buildings in Tokyo held by Japanese Real Estate Investment Trusts (J-REITs) and appraised by professionals from Japan's largest real estate appraisal organisation. Multiple regression analysis is employed to compare property valuation models with and without façade sensory evaluation variables, while path analysis is used to identify the indirect effects of façade-related components, such as building colour, on property value. Findings: Façade sensory evaluation significantly improves the explanatory power of property valuation models and influences property values at the 1% significance level. Among building attributes, the number of floors has the strongest effect on façade evaluation, followed by colour saturation, building age and brightness. Meanwhile, building colour indirectly affects property value through appraisers' façade evaluations. Practical implications: By quantifying appraisers' tacit knowledge of façade impressions, this study clarifies sources of information asymmetry in property markets and highlights the importance of perceived building scale and colour in professional office property valuation. Originality/value: Through path analysis that incorporates building colour, this study uniquely demonstrates how real estate appraisers' façade impressions structurally influence property value. In doing so, it extends existing research on façade design and property valuation.
- 6 June 2026
- Published by Springer Nature
- p. 230-244
Forecasting Residential Gas Demand in an Aging Regional City Using a Hybrid Microsimulation Model
- 6 June 2026
- Published by Springer Nature
- p. 230-244
Forecasting Residential Gas Demand in an Aging Regional City Using a Hybrid Microsimulation Model
- 18 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
Data-Driven Group Persona Generation from POS Data Using Generative AI Prompts
This study proposes a methodology for constructing data-driven ‘group personas’ using generative AI prompts, leveraging over 240,000 POS data from a restaurant located in Ginza, Japan. The analysis suggests that persona estimation is feasible at the group level. The findings of this study demonstrate the potential for formulating effective marketing strategies and developing new menus targeted at multi-guest visits in the food service industry, thereby contributing to the advancement of datadriven decision-making.
- 18 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
Data-Driven Group Persona Generation from POS Data Using Generative AI Prompts
This study proposes a methodology for constructing data-driven ‘group personas’ using generative AI prompts, leveraging over 240,000 POS data from a restaurant located in Ginza, Japan. The analysis suggests that persona estimation is feasible at the group level. The findings of this study demonstrate the potential for formulating effective marketing strategies and developing new menus targeted at multi-guest visits in the food service industry, thereby contributing to the advancement of datadriven decision-making.
- 5 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Evaluating LLM-Induced Investment Biases via Persona Design and Prompt Analysis: A Quantitative Study of Post-Retirement Asset Depletion Risk Through Social Simulation
This study quantitatively evaluates whether responses from large language models (LLMs) can reinforce individual investment biases, and how such reinforcement may alter medium- to long-term asset depletion risks, through a three-stage methodology involving persona design, prompt analysis, and social simulation. First, representative personas are constructed by clustering data from a large-scale survey, incorporating plausible individual profiles. For each persona, financial consultations are conducted with an LLM using multiple prompts that integrate varying risk preferences. The risk tolerance exhibited in the LLM responses is scored and statistically analyzed for significant differences via one-way ANOVA. Next, asset allocations derived from these risk preferences are input into Monte Carlo simulations-including exogenous macroeconomic variables-to estimate age-specific depletion rates. The results demonstrate that changes in the proportion of risky assets can affect the probability of future asset depletion. This approach is novel in that it empirically links the impact of prompt design on long-term household risk by path wise visualization of response bias stemming from the “sycophancy problem” of LLMs.
- 5 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Evaluating LLM-Induced Investment Biases via Persona Design and Prompt Analysis: A Quantitative Study of Post-Retirement Asset Depletion Risk Through Social Simulation
This study quantitatively evaluates whether responses from large language models (LLMs) can reinforce individual investment biases, and how such reinforcement may alter medium- to long-term asset depletion risks, through a three-stage methodology involving persona design, prompt analysis, and social simulation. First, representative personas are constructed by clustering data from a large-scale survey, incorporating plausible individual profiles. For each persona, financial consultations are conducted with an LLM using multiple prompts that integrate varying risk preferences. The risk tolerance exhibited in the LLM responses is scored and statistically analyzed for significant differences via one-way ANOVA. Next, asset allocations derived from these risk preferences are input into Monte Carlo simulations-including exogenous macroeconomic variables-to estimate age-specific depletion rates. The results demonstrate that changes in the proportion of risky assets can affect the probability of future asset depletion. This approach is novel in that it empirically links the impact of prompt design on long-term household risk by path wise visualization of response bias stemming from the “sycophancy problem” of LLMs.
- 10 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 197-202
Visualization and Interpretation of Technology Lifecycles from Patent Data: The S-Curve as an XAI Tool
This study attempts to clarify the characteristics of technology lifecycles by analyzing patent data using a Large Language Model (LLM). This analysis focuses on 14,837 international patents related to blockchain technology filed between January 2014 and July 2025. The methodology includes patent classification using LLM and clustering, and visualization of the results, to explore the relationship with market valuation. The analysis revealed that 1. there are eight major technical clusters of blockchain identified using Sentence BERT and K-means, and 2. S-curve visualization clearly delineates the characteristics of each cluster. This study not only presents interesting findings regarding blockchain technology but also demonstrates that technology trend analysis based on patent data is a feasible application area for LLMs and Explainable AI (XAI).
- 10 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 197-202
Visualization and Interpretation of Technology Lifecycles from Patent Data: The S-Curve as an XAI Tool
This study attempts to clarify the characteristics of technology lifecycles by analyzing patent data using a Large Language Model (LLM). This analysis focuses on 14,837 international patents related to blockchain technology filed between January 2014 and July 2025. The methodology includes patent classification using LLM and clustering, and visualization of the results, to explore the relationship with market valuation. The analysis revealed that 1. there are eight major technical clusters of blockchain identified using Sentence BERT and K-means, and 2. S-curve visualization clearly delineates the characteristics of each cluster. This study not only presents interesting findings regarding blockchain technology but also demonstrates that technology trend analysis based on patent data is a feasible application area for LLMs and Explainable AI (XAI).