By Jia Nan Song
On May 8, the Central Cyberspace Affairs Commission announced that since the release of the “Negative List of Algorithmic Governance for Life Service Platforms (Trial)” in January, major platforms such as Meituan, Taobao, JD.com, Didi, and Baidu have actively responded and implemented optimization measures, with a total of 63 optimization measures, 139 commitments to comply with algorithm requirements, and 125 measures to be promoted within a time limit, achieving initial results in improving algorithm transparency, fairness, and rationality.
The optimization of platform algorithms mainly involves seven aspects: order allocation algorithms, time estimation algorithms, safety guarantee algorithms, income extraction algorithms, pricing algorithms, complaint handling, and algorithm transparency. For example, Meituan’s food delivery service will explore reserving at least 15 minutes of delivery time for riders in some scenarios; Didi will reduce the upper limit of commission from 29% to 27% and launch the “Commission Return” feature to automatically return the excess commission.
All major platforms have promised not to set different prices for the same goods or services under the same transaction conditions based on consumers’ browsing records, payment intentions, payment abilities, consumption records, and consumption preferences.
For a long time, the “algorithm-first” logic has led platforms to worship efficiency: riders risk reversing traffic to avoid fines, drivers suffer from opaque commissions, and consumers are killed by big data, making algorithms a tool for platforms to unilaterally dominate the distribution of interests.
According to QuestMobile data, as of July last year, the number of monthly active users of online car-hailing drivers reached 29.24 million, a year-on-year increase of 23.3%. Among them, Didi’s car owners had 20.79 million monthly active users. On the other hand, the number of employees in the express delivery and food delivery industries also reached approximately 20 million, with 14.035 million food delivery personnel and 5.394 million express delivery personnel.
The protection of the rights and interests of these new employment groups is directly related to the well-being and social stability of the people. The essence of the optimization of platform algorithms is the transformation of algorithms from “efficiency tools” to “public governance rules”.
However, the implementation of rules is difficult. The real test of algorithm optimization lies in the daily execution details, and the core issue will shift from “how to optimize algorithms” to “how to prevent optimized algorithms from going astray”.
A relevant person in charge of the Central Cyberspace Affairs Commission said that some platforms still have problems such as “selective rectification” and “not rectifying unless others do”, which is far from the expectations of the masses and the new employment groups.
Some platforms’ “pseudo-rectification” tactics have already appeared: some have cancelled fines for overtime but have reduced delivery fees to exert pressure; some have publicized algorithm logic but it is obscure and difficult to understand; and some have selectively executed regulatory priorities, ignoring user complaints about hidden problems. More hidden is the “algorithm rebound” phenomenon. A driver reflected that although the platform has reduced the upper limit of commission, the distribution of peak hours is still opaque, and the actual income has not increased significantly.
The root cause of these problems lies in the deep-seated contradiction between the platform’s profit logic and the protection of laborers’ rights and interests: low commission and strong protection mean increased operating costs, and some platforms lack the internal motivation to execute long-term, and are prone to “reverting to their old ways” when regulatory pressure weakens.
To break the execution dilemma, platforms must take proactive measures beyond “passive compliance”. First, they must establish a “perceivable and participatory” algorithm negotiation mechanism, allowing riders and drivers to become active participants in the rules. For example, some food delivery platforms have established a negotiation system, which is more effective than cold system notifications. Second, they must strengthen technical guarantees to prevent “human intervention” in algorithm execution. Finally, they must break the “algorithm black box” and dissolve the trust crisis through transparency.
New employment groups should make good use of the feedback channels provided by platforms, expressing their demands through regular channels. When encountering unreasonable order allocation, opaque commission, and other issues, they can report them through the APP feedback entrance, union hotline, or algorithm consultant committee, submitting specific order information and screenshots to make optimization suggestions more targeted.
At the same time, practitioners should rationally view algorithm optimization, understanding that mandatory rest is not a restriction on income but a guarantee of long-term work ability; cancelling fines for overtime is not a relaxation of requirements but a return to the “safety first” principle. This kind of bilateral interaction can make algorithms continuously improve in practice, truly achieving a win-win situation for all parties.
The Central Cyberspace Affairs Commission has made it clear that the next step will be to organize special inspections and deal with serious problems according to the law, which means that algorithm governance will not be a “flash in the pan”.
For the entire platform economy, it is necessary to establish a “platform self-inspection + third-party audit + public supervision” three-dimensional regulatory system in the future. For example, introducing independent institutions to evaluate algorithm execution and publishing audit results; unblocking the 12377 network reporting channel, encouraging consumers and practitioners to report violations; and linking algorithm execution with platform credit ratings to form a long-term constraint. Only in this way can the “wait-and-see” mentality of platforms be eliminated, forcing them to internalize algorithm optimization as their core competitiveness.
From the release of rules to thorough implementation, algorithm governance still has a long way to go. It is worth noting that “pseudo-rectification” may reduce short-term costs, but it will ultimately undermine the platform’s long-term credibility and be abandoned by the market.
Algorithm optimization is not a zero-sum game, but a necessary path for the sustainable development of platforms. Compliant platforms not only win social reputation but also achieve business value growth. When riders no longer have to compete with time, drivers no longer have to worry about commission, and consumers no longer have to worry about big data killing, platforms can build a more stable trust ecosystem. The user stickiness and brand premium brought by this trust are far more valuable than short-term efficiency exploitation.
