Nature Commentary: Human-AI Complementarity Should Be Achieved Through Augmentation, Not Simulation
A commentary published in Nature Reviews Psychology on February 9, 2026, questions the necessity of “Cognitive AI” for achieving human-AI complementarity.The article points out that joint human-AI decision-making can achieve superior results, but the path to this should focus on augmentation rather than simulating human cognitive processes. This viewpoint challenges the “Cognitive AI” framework previously proposed by Gonzalez and Heidari.
Background of the Proposal in Question
In a Perspective article, Gonzalez and Heidari proposed that “Cognitive AI,” by simulating the information processing systems of the human mind, can aid human-AI complementarity in dynamic decision-making. This proposal distinguishes Cognitive AI from traditional data-driven AI, emphasizing that the former interprets and replicates human decision-making processes based on formal cognitive models.
Zhicheng Lin’s commentary raises objections to this framework, arguing that it is built on two debatable premises, which affects the effectiveness of achieving human-AI complementarity.
Questioning the Two Core Premises
The article first questions the substantive distinction between Cognitive AI and data-driven AI. The author notes that data-driven systems like large language models, trained on human-generated data including behavioral data, can already display emergent human-like properties, such as biases. These properties arise from implicit modeling of the statistical regularities in human decisions, without relying on formal cognitive models.
Second, the article challenges the assumption that an effective human-AI team requires the AI to replicate human cognitive processes. The author believes this premise overlooks the fact that data-driven methods can already achieve similar design principles, including transparency, adaptability, and personalization.
Rationale for Augmentation Over Simulation
The commentary emphasizes that human-AI complementarity is better achieved through augmentation rather than simulation. Data-driven AI can effectively support complementary collaboration without strictly aligning with formal cognitive models.
The author points out that the integration of Cognitive AI and data-driven methods lacks a clear boundary in practice, and the latter has already demonstrated sufficient capabilities in many aspects to support complementarity. This perspective offers an alternative angle for human-AI collaboration design, focusing on using AI to augment human abilities to enhance overall decision quality.
Conclusion and Implications
Zhicheng Lin concludes that human-AI complementarity requires augmentation, not simulation. This stance is based on observations of existing AI capabilities, indicating that data-driven systems have already implicitly incorporated features of human decision-making.
Although the article does not provide new experimental data, through theoretical analysis and citation of existing research, it provides a reference direction for the design of human-AI teams in dynamic decision-making scenarios.
