Author(s): Taylor Onate Egerton and Davies Isobo Nelson

Abstract: Algorithmic decision support systems are transforming global financial markets, yet fully automated trading often fails under non-stationary or rare market conditions, propagates biases and undermining transparency, trust, and regulatory acceptance. This paper introduces a Selective Human-Machine Integration Framework (SHMIF) that enhances short-term trading decision quality for equities listed on the Nigerian Stock Exchange (NSE) by strategically combining machine intelligence with human expertise through selective routing, adaptive explain ability, and continuous feedback-driven learning. The proposed SHMIF architecture comprises four core modules. Further, the study employed a controlled pilot experiment involving three professional NSE analysts to evaluated 3,600 decision trials across human-only, machine-only, and hybrid configurations using data from 2019-2024, covering 25 liquid NSE equities. Experimental results show that SHMIF achieved 91.3% decision accuracy, outperforming human-only 77.9% and machine-only 83.4% baselines, yielding a 13.4% and 7.9% improvement respectively. The framework produced a 32.6% increase in profitability, reduced volatility (10.9% vs. 12.3%), and enhanced risk-adjusted returns (Sharpe ratio: 1.45 vs. 1.19). Trust ratings averaged 4.8/5, while only 43% of cases required human intervention. Statistical analysis confirmed significant effects of decision mode (F (2,22) =26.41, p<0.001) and explanation type (F (2,22) =18.93, p<0.001), with case-based reasoning yielding the highest accuracy (92.1%) and trust. The Hybrid Complementarity Index (21.7%) indicates strong synergy between human and machine intelligence. These findings demonstrate that selective human-machine integration substantially improves decision quality, interpretability, and trustworthiness in financial markets. The SHMIF framework provides a scalable blueprint for responsible AI deployment in emerging financial markets, supporting the transition from automation to collaborative intelligence in complex decision environments.

DOI: 10.33545/27076571.2025.v6.i2d.211

Pages: 316-322 | Views: 73 | Downloads: 40

Download Full Article: Click Here