By Wu Jing, Professor of Philosophy, East China Normal University
With the growing prevalence of digital intelligence technology and the rapid expansion of the digital economy, algorithms have effectively become the "invisible hand" behind digitalization. As research in the humanities and social sciences deepens, critical voices against algorithms have grown increasingly strong. However, a one-sided understanding of algorithms risks falling into the trap of "algorithmic stigmatization"—reducing algorithms to mere tools of control, making them scapegoats for all problems in the digital age, overlooking multiple other possible factors behind digital intelligence risks and injustices, and rigidly fixing the current application of algorithms as their eternal nature, thereby denying their functionality and potential for improvement.
In contrast to the critical trend that blames algorithms for all negative effects of digital applications, the field of algorithms has in recent years demonstrated remarkable room for improvement. This progress is reflected not only in enhanced technical performance but also in the realization of value-driven goals and the refinement of governance mechanisms. Technically speaking, iterative optimization of algorithms has shown multiple positive trends. At the value level, the embedding of diverse perspectives is receiving increasing attention. Early algorithmic optimization often pursued a single goal, which easily led to value biases. Contemporary algorithmic design, however, has begun exploring multi-objective optimization, value-sensitive design, ethical impact assessments, and other approaches, seeking to involve multiple stakeholders early in the development process. Of course, algorithmic governance cannot rely solely on technical self-improvement; it requires an institutional regulatory framework to ensure that algorithms are used correctly, well, and in service of people. Thus, governance must focus on two levels simultaneously: technically, it must promote the standardization and third-party implementation of algorithmic auditing, establishing benchmark datasets for performance testing and fairness evaluation; institutionally, it must improve the legal framework for algorithmic registration, impact assessment, and accountability, clarifying platform companies' obligations in algorithmic governance.
It is worth emphasizing that the improvability of algorithms does not mean that all problems can be solved through technical means. Some value conflicts are fundamentally irreconcilable, and some social inequalities require structural changes beyond the technical realm. Nevertheless, recognizing the improvability of algorithms means rejecting the reductionist fallacy that "algorithms are the root of all evil," maintaining a positive expectation for the potential of technical optimization, and exploring the optimal integration of technical improvements and institutional reforms in specific contexts.
In a digital intelligence society, algorithms possess real power attributes—power that must be scrutinized, constrained, and held accountable. They also have genuine limitations and boundaries, which should likewise be acknowledged and addressed. Facing the challenges of the digital intelligence era, the unique contribution of the humanities and social sciences lies precisely in delving into the inner logic of technological operations, revealing their complex interactions with social systems, and providing value guidance and institutional frameworks for technological improvement. This means that scholars must maintain both sensitivity and vigilance toward technological power while recognizing the possibility and necessity of technological optimization; they must uphold the ultimate concerns of humanistic values while avoiding the absolutization of the opposition between values and technology.
Published on May 7, 2026