Preserving Human Freedom in the Information Age
Bill Hibbard May 2020
"Whoever would overthrow the liberty of a nation must begin by subduing the freeness of speech." - Benjamin FranklinSixty-three years ago, in The Hidden Persuaders, Vance Packard described efforts to control our thoughts covertly. Now artificial intelligence (AI) is advancing to the point where it can model the mental processes of each human to find the best ways to persuade them, and model social processes to find the best ways to coordinate its persuasive messages to humans. Facebook, Instagram, Twitter, and Snapchat demonstrate the power of peer pressure via electronic media. China gives us an example of how a central political authority can use AI, connected via the Internet to all the devices of people's daily lives, to surveil and control their population. AI that surpasses the human mind and is intimately woven into society will have power to control public opinion unlike anything we see today.
Such power is tempting to political leaders who want to advance their agendas, and to ordinary citizens frustrated that not everyone agrees with their views on climate change, social justice, human freedom, etc. There in an irresistible attraction between politics and AI, a marriage made in heaven or perhaps hell. The deep division in our current politics reflects the depth of persuasion within each faction. Inevitably the persuasive power of AI will be employed to control the consensus of human opinion and hence control society. However, this power should be based on open rather than hidden information. I certainly think Covid-19 is real but imagine a government that concocted a story of a fake pandemic and then staged a break-down of person-to-person electronic communications, to stop social interactions and isolate people. The danger in that case comes from hiding the information that the pandemic and communications break-down are fake. Conspiracy theories thrive when there is a lack of transparency and accountability by those with power over our lives.
My purpose here is to advocate for a set of simple policies for human use of information processing equipment (IPE), to ensure that human opinions are formed from open information and that no relevant information is hidden. IPE includes any equipment for communicating or processing information, such as phones, computers, TVs, the Internet, and large AI servers. The proposed policies are:
Freedom of speech. People must be free to provide any speech and other inputs to IPE, including encrypted input. They may face legal liability for certain inputs to IPE.
Privacy. People must be able to limit the scope of transmission and effect of their speech and other inputs to IPE. Possible scopes include a) self, b) a well-defined group of humans, and c) all humans. This policy applies to data about humans generated incidentally, such as cell phone tracking. There may be exceptions for information mandated to be public, such as court records. When scope is limited to self, humans may direct IPE to forget inputs within a well-defined time interval.
Accountability. People must be accountable for all their inputs to IPE (i.e., no anonymous speech). The input information and the name of its human source must be available to any person within the information's scope of effect. If the source of information is a corporation then the information, the name of the corporation, and the names of all human corporate officers and major human owners, including through chains of corporate ownership, must be available to any person within the information's scope of effect. People should not be held accountable for incidental data about them, generated unintentionally. For example, people should not generally be accountable for data about them used to compile population statistics, but the person whose input causes IPE to compile population statistics should be accountable to anyone in the scope of effect of the statistics.
Transparency. Every person must have access to all source code and design documents for all IPE that affect them or have information about them (transparency may be viewed as a special case of accountability, where source code and designs are inputs to IPE). There may be exceptions to transparency and accountability for military IPE providing that the IPE does not exchange information with any humans other than uniformed military personnel. Most people will not be able to make effective use of source code and design documents, but some will be able to and the precedent of Internet security testing indicates that they will provide the public with simple explanations and with software to analyze how IPE affects them. Transparency of all IPE will expedite enforcement of patents and copyrights.
Liability. People may face legal liability for their speech and other inputs to IPE, with scope beyond self. This may include advocating or knowingly causing violence, defaming or libeling other humans, sharing information about another person beyond the intended scope (e.g., revenge porn), and child porn (in this case, even when the scope is self). Liability must be determined by a human legal process rather than by IPE.
Mechanisms can be built into IPE for implementing the privacy, accountability, and transparency policies. IPE can alert human prosecutors to potential liability.
The proposed transparency and accountability policies represent a radical change from current practice. Corporations resist calls for transparency of their IPE and major human owners of corporations may resist making their names available to humans affected by inputs to IPE from the corporations. However, the potential of AI to totally transform human life justifies radically new policies for corporate governance. A recent AI policy document called for third party auditing of AI rather than public transparency. With AI the high value of hidden information will create an unacceptable risk of conflict of interest for any third party. Only open transparency and accountability can be trusted to serve the public interest.
It is worth remembering that the proposed policies apply only to IPE and that the best things in life have nothing to do with IPE. Let's not let IPE ruin those.