|
Finally, a newsletter that does NOT focus on Donald Trump’s administration. I am so tired of being in the blast zone of his behavior and feeling I have to write about it in every newsletter! Today, I am pondering an unresolved but critical ethical dilemma of the internet era, made worse by the coming of artificial intelligence: are internet platforms (and now AI models available to the public) responsible for the content users post on them or create with them? Meta, Google, and other platform companies have relied on Section 230 of the Communications Decency Act of 1996 to shield them from liability for content posted by their users. In short, Section 230 declares that the platforms are not “publishers,” which would trigger responsibility for the posted content. Conversely, the law also provides a safe harbor for platforms to moderate and remove content it deems obscene, violent, harassing, or otherwise objectionable. When Section 230 was written in 1996, the internet and the platforms were in their early years. Arguments that treating platforms as publishers would drastically limit their adoption and usefulness. Since 1996, there has been a growing awareness that shielding platforms created a Wild West environment where huge volumes of troubling content have spread. Proposals for reform or repeal of Section 230 have, to date, all failed. Some argue that the platforms have too much immunity and that the public damage being done by harmful postings (hate speech, misinformation, etc.) is too great and must be controlled. Others, echoing arguments in 1996, contend that any change in Section 230 would destroy the platforms and the good content posted there. There is also a debate over the safe harbor given to platforms to moderate content, if they choose to do so. In an age of disinformation and misinformation, Republican politicians and Trump supporters (wow! I can’t keep him out completely) argue that many platforms censor only conservative voices. As platforms have matured, the algorithms created by the platforms themselves to serve up “related” content to users have been challenged. A growing number of observers have argued that such algorithms reinforce extreme content (leading users down “rabbit holes”) and make the platforms “addictive”—designed to capture and manipulate users to spend excessive time on the platforms. Two recent developments dramatize this last concern. Since 2022, a bipartisan bill, the Kids Online Safety Act (KOSA), has been repeatedly debated in Congress. Designed to impose a “duty to care” on platforms to prevent harms to minors (suicide, substance abuse, eating disorders, and sexual exploitation), the bill has made little progress toward adoption. It has passed Senate committees but has not reached a full vote in either the House or the Senate. In the absence of federal action, some states such as California have passed related child protection laws that focus on the collection of child data, reducing addiction, and relieving mental health effects. Platforms have opposed most of these laws, arguing that young users can fake their ages and work around restrictions. The other recent development is a California civil liability lawsuit, K.G.M. v. Meta et al, which alleges social media platforms were designed to be addictive and harmful to young people. Two platforms, TikTok and Snapchat, settled out of court early, but the trial is now ongoing in Los Angeles Superior Court with Meta (Instagram and Facebook) and YouTube (Google) as defendants. Plaintiffs must convince the court that the companies intentionally built their platforms to hook individuals, particularly the young, and were aware that the platforms harmed vulnerable users, prioritizing engagement and profits above safety. The trial began on February 9, and Mark Zuckerberg, the CEO of Meta, testified last week. In several hours on the stand, Zuckerberg argued that the platforms were not intentionally designed to be addictive to minors, but instead, to help all users find content they valued. He argued there were no corporate goals based on children’s time on the platforms, and that company metrics tracking user time were considered proxies for user satisfaction with the platform, not for their addiction. He also argued there is no conclusive evidence that use of the platforms causes users’ mental health problems. Finally, he argued that the platforms are doing a good job of limiting use by children under 13 years old and of heading off lying about one’s age. There is no question that this is a thorny or “wicked” ethical question, with substantive arguments on both sides. Imposing a “duty to care” would limit the expansion and use of the platforms, just as they are adding expensive artificial intelligence to their features. But the damages being done to children, while still hard to prove scientifically and certainly hard to measure, are so great as to make caution an ethical obligation. In two other major business ethics issues of the past 50 years—the sale of tobacco products and the continued extraction and promotion of oil and gas challenged the most fundamental values of the companies involved. Some companies fully acknowledged the ethical and policy dilemmas, taking leadership to identify solutions. Others, R.J. Reynolds in tobacco and Exxon Mobil in the oil industry, in my opinion, have taken the extreme position that there is no proven harm to humans and that regulation would compromise the free enterprise system. By contrast, at one point, British Petroleum, led by one of my former students, Lord John Browne, embraced alternative sources of energy and branded itself as Beyond Petroleum (BP). Some argue that Meta is today the most resistant of the platform companies in the current struggle. The responsible company, I would argue, is in the best position to know the harms and risks of its products, and therefore to help to resolve such great business ethics dilemmas. The responsible company ought to be engaged in ongoing and independent research on harms and risks—and in the search for balanced solutions to the issues at hand. However, just as politics has become more partisan, the solution of any significant business ethics dilemma is hampered by extreme positions on all sides, often magnified by online platforms. The temptation to hide or obfuscate emerging evidence of harm—and to resist solutions—is strong, and both tobacco and oil companies gave in to that temptation. Let’s hope that in 20 years, we are not lumping Meta into the gallery of business shame along with bad actors in tobacco, fossil fuels, and opioid industries. Here is perhaps an obvious prediction: the next great business ethics struggle in our economic history will be how artificial intelligence is deployed. Companies are already differentiating themselves in how they are approaching the ethics risks. Last Friday, we saw Anthropic, which has taken a more cautious—and some would say responsible—approach to the use of its Claude LLM, banned by Defense Secretary Pete Hegseth as a federal contractor. Anthropic would not agree to unrestricted use of its AI model by the Defense Department, including use for mass domestic surveillance and completely autonomous weapons. OpenAI, Elon Musk’s xAI, and other AI pioneers indicated their eagerness to replace Anthropic’s Claude with no restrictions. Hmmmm.
|
