Artificial intelligence for socially responsible investment: a multi-objective decision-making framework

Abstract This paper presents a novel multi-objetive methodology for optimizing financial investment decisions. To do this, we integrate machine learning (ML) techniques into a Black–Litterman (BL) model, using conditional value at risk (CVaR) as a risk measure within a socially responsible investment framework. The proposed approach employs ML models, including long short-term memory, random forest, artificial neural networks, gated recurrent unit, and autoregressive integrated moving average (ARIMA), to forecast asset returns. These predictions are aggregated through a model based on historical performance, and they are incorporated into a Black–Litterman–CVaR model (BL–ML–CVaR). Environmental, social, and governance (ESG) criteria are used to construct portfolios that align financial returns with sustainable investment practices. Results demonstrate that the BL–ML–CVaR portfolio offers higher returns and balanced risk compared to traditional models, with the ESG-integrated version achieving competitive performance while adhering to sustainability goals. This study provides valuable insights into how ML-driven models can enhance investment strategies in some emerging sectors facing environmental challenges, offering a path toward more sustainable and responsible financial practices.