算法规制:作为治理工具的机器学习

Algorithmic Regulation:Machine Learning as a Governance Tool

算法规制,有助于行政机关作出更智能、更快速和更一致的决定,从而在给社会带来更好结果的同时降低行政成本。算法规制的不同之处,源于机器学习的不透明性和相对自主性。同任何新型的政策或工具一样,算法规制也引发了一系列担忧,包括可问责性、平等、正当程序、隐私、透明度和权力滥用等。除上述挑战外,机器学习技术还需要面对其他限制,如价值的完整性和精确性。然而,对算法规制最严重的限制将非源自技术本身,而是来自使用这些工具的人。为负责任地使用算法,行政机关应当正视技术的风险和局限,以及在公共部门内培养部署算法工具的人力能力。

Algorithmic regulation helps to make smarter,faster and more consistent decisions,thereby facilitating outcomes that are better for society and potentially of lower cost to government.The difference of algorithm regulation stems from the opacity and relative autonomy of machine learning.As with any new type of policy or tool,algorithmic regulation also raises a series of concerns,including accountability,equality,due process,privacy,transparency and abuse of power.In addition to the above challenges,machine learning technology also needs to face other limitations,such as ensuring value completeness and value precision.However,the most serious limitations on algorithmic regulation will stem not so much from the technology itself but from the humans that seek to use these new tools.In order to use algorithms responsibly,governments need to approach the prospects of algorithmic regulation with eyes wide open to the technology's risks and limitations,and they should prepare to develop the human capabilities within the public sector.