Chao on Real Hypothetical Negotiations

Bernard Chao has posted a paper on ssrn titled Real Hypothetical Negotiations.  Here is a link to the paper, and here is the abstract:

Patent law's prevailing method for awarding reasonable royalties relies on a "hypothetical negotiation" framework that asks what royalty a willing licensor and willing licensee would have agreed to. In practice, this approach has devolved into a battle of paid experts who manipulate the same evidence to reach dramatically different conclusions, often diverging by orders of magnitude. The current system's reliance on the unwieldy fifteen-factor Georgia-Pacific test, combined with inadequate judicial oversight and jury decision-making limitations, has transformed what should be a reliable proxy of market behavior into a stylized fiction used for strategic advocacy.

This article proposes and pilots the first empirically grounded alternative to expert driven reasonable royalty determinations. Drawing on patent law, experimental jurisprudence, negotiation theory and generative AI decisionmaking, it introduces "real hypothetical negotiations." Under this proposal, royalty rates are determined through simulated negotiation. Departing from the traditional remedial goal of strictly restoring the patentee to a pre-infringement state, this proposal prioritizes the constitutional mandate of promoting innovation. Accordingly, the briefing materials include both pre-infringement and post-infringement information to best calibrate damages to the invention’s actual economic value. 

To test this approach, the article reports results from two pilot studies based on Summit 6 LLC v. Samsung Electronics Co., where opposing damages experts reached dramatically different conclusions. Law students using the same case materials reached settlement in only two out seven of negotiations, while competing AI large language models (LLMs) using identical materials achieved settlement in all six cases. These mixed results illustrate both the challenges and promise of operationalizing this approach.

The article also describes several potential enhancements to the basic proposal: 1) blinding experts to the party that retained them, or alternatively, 2) court-appointed neutral experts to oversee the negotiation process, 3) conducting simulations multiple times to address outlier results and improve statistical reliability, and 4) using LLMs themselves as negotiators. The article concludes that real hypothetical negotiations offer valuable potential for both actual cases and controlled experiments. Future experiments could filter out value unrelated to innovation, revealing the patent’s true technical contribution and guiding damages awards toward outcomes that advance innovation.

This is a very interesting paper.  I agree with Professor Chao's assessment that "[t]he combination of vague guidance, inadequate judicial oversight, and jury limitations creates a system where reasonable royalty determinations are frequently divorced from economic reality," and I find his proposals for reform intriguing (though I suspect it might be a heavy lift to get them introduced into U.S. practice).  Still, the use of AI as a research tool to assist in negotiations could well prove beneficial, notwithstanding its potential limitations as noted in the article.