What AI Brings to Financial Investment: Limitations of Recommendation Models Compared to Live‑Streaming Commerce

During the pandemic, live‑streaming e‑commerce in China surged, with the industry reaching 433.8 billion CNY in 2019 and an expected user base of 524 million in 2020, prompting a comparison between this model and traditional recommendation algorithms.

The article focuses on a specific live‑streaming pattern—content seeding on platforms such as Xiaohongshu, Douyin, Kuaishou, and self‑media, which drives traffic to Taobao and Tmall, exemplified by influencers like Li Jiaqi and Luo Yonghao who leverage personal trust and charisma to convert viewers into buyers.

While trust lowers the cost of choice for consumers, recommendation models in e‑commerce and video platforms are evaluated mainly by short‑term metrics such as accuracy, recall, click‑through rate, and conversion rate, often through A/B testing frameworks.

Standard recommendation‑system evaluation metrics are listed: accuracy (precision), recall, user satisfaction (click‑through, conversion), coverage, diversity, novelty, explainability, and robustness. These metrics generally align with user interests but the ultimate decision to deploy a model is dictated by the platform’s commercial objectives—transaction volume for e‑commerce, or engagement time for media platforms.

The article notes that big‑data user‑behavior feeds recommendation models, which are theoretically grounded in revealed‑preference theory, yet real‑world deployment prioritizes platform profit over true user utility, especially when user preferences are shown to be stochastic according to economic and neuroscientific research.

Four key shortcomings of recommendation models versus live‑streaming are identified: lack of a trust anchor and professional credibility, overly short‑term evaluation criteria, commercial bias in model rollout, and a focus on predicting likely user actions rather than guiding more rational decisions.

Quoting Steve Jobs on Siri’s AI positioning, the piece suggests that recommendation systems will evolve toward intelligent assistants, and it concludes with a promotion of an upcoming book on AI in financial investment, the author’s credentials, and several endorsements from industry leaders.