An important step in the development of value alignment (VA) systems in AI is understanding how VA can reflect valid ethical principles. We propose that designers of VA systems incorporate ethics by utilizing a hybrid approach in which both ethical reasoning and empirical observation play a role. This, we argue, avoids committing the "naturalistic fallacy," which is an attempt to derive "ought" from "is," and it provides a more adequate form of ethical reasoning when the fallacy is not committed. Using quantified model logic, we precisely formulate principles derived from deontological ethics and show how they imply particular "test propositions" for any given action plan in an AI rule base. The action plan is ethical only if the test proposition is empirically true, a judgment that is made on the basis of empirical VA. This permits empirical VA to integrate seamlessly with independently justified ethical principles.

Training datasets for machine learning often have some form of missingness. For example, to learn a model for deciding whom to give a loan, the available training data includes individuals who were given a loan in the past, but not those who were not. This missingness, if ignored, nullifies any fairness guarantee of the training procedure when the model is deployed. Using causal graphs, we characterize the missingness mechanisms in different real-world scenarios. We show conditions under which various distributions, used in popular fairness algorithms, can or can not be recovered from the training data. Our theoretical results imply that many of these algorithms can not guarantee fairness in practice. Modeling missingness also helps to identify correct design principles for fair algorithms. For example, in multi-stage settings where decisions are made in multiple screening rounds, we use our framework to derive the minimal distributions required to design a fair algorithm. Our proposed algorithm decentralizes the decision-making process and still achieves similar performance to the optimal algorithm that requires centralization and non-recoverable distributions.

Branches of Co-op in the south of England have been using real-time facial recognition cameras to scan shoppers entering stores. This story originally appeared on WIRED UK. In total 18 shops from the Southern Co-op franchise have been using the technology in an effort to reduce shoplifting and abuse against staff. As a result of the trials, other regional Co-Op franchises are now believed to be trialing facial recognition systems. Use of facial recognition by police forces has been controversial, with the Court of Appeal ruling parts of its use to be unlawful earlier this year.

Self-reinforcing feedback loops are both cause and effect of over and/or under-presentation of some content in interactive recommender systems. This leads to erroneous user preference estimates, namely, overestimation of over-presented content while violating the right to be presented of each alternative, contrary of which we define as a fair system. We consider two models that explicitly incorporate, or ignore the systematic and limited exposure to alternatives. By simulations, we demonstrate that ignoring the systematic presentations overestimates promoted options and underestimates censored alternatives. Simply conditioning on the limited exposure is a remedy for these biases.

DJI hasn't been deterred by the US Commerce Department's trade ban. The drone maker told TechCrunch that Americans can buy and use its products "normally" despite the company's presence on an entity list barring US companies from doing business with the firm. DJI "remains committed" to making innovative hardware, a spokesperson said. The Commerce Department added DJI to the list for having allegedly "enabled wide-scale human rights abuses" in China, including drones used to help with the surveillance and persecution of Uyghur Muslims. It's not certain how long usual business might last.

For Sonos, 2020 began in dramatic fashion. While the tech world was focused on CES, the company made a splash by suing Google for allegedly infringing five of its wireless speaker patents. Sonos said this was just a small portion of Google's overall infractions, noting that both Amazon and Google likely violated about 100 patents each. Google counter-sued in June, and Sonos filed more charges in September. Sonos is well within its rights to defend its patent portfolio -- and the company has been working on wireless music-streaming tech for longer than just about anyone, so it's entirely possible its claims have merit.

Hate speech is a challenging issue plaguing the online social media. While better models for hate speech detection are continuously being developed, there is little research on the bias and interpretability aspects of hate speech. In this paper, we introduce HateXplain, the first benchmark hate speech dataset covering multiple aspects of the issue. Each post in our dataset is annotated from three different perspectives: the basic, commonly used 3-class classification (i.e., hate, offensive or normal), the target community (i.e., the community that has been the victim of hate speech/offensive speech in the post), and the rationales, i.e., the portions of the post on which their labelling decision (as hate, offensive or normal) is based. We utilize existing state-of-the-art models and observe that even models that perform very well in classification do not score high on explainability metrics like model plausibility and faithfulness. We also observe that models, which utilize the human rationales for training, perform better in reducing unintended bias towards target communities. We have made our code and dataset public at https://github.com/punyajoy/HateXplain

Recent papers in explainable AI have made a compelling case for counterfactual modes of explanation. While counterfactual explanations appear to be extremely effective in some instances, they are formally equivalent to adversarial examples. This presents an apparent paradox for explainability researchers: if these two procedures are formally equivalent, what accounts for the explanatory divide apparent between counterfactual explanations and adversarial examples? We resolve this paradox by placing emphasis back on the semantics of counterfactual expressions. Producing satisfactory explanations for deep learning systems will require that we find ways to interpret the semantics of hidden layer representations in deep neural networks.

The latest letter doesn't tie directly to the Algorithmic Accountability Act, but it is part of the same move by certain congressional members to craft legislation that would mitigate AI bias and the other harms of data-driven, automated systems. Notably, it comes amid mounting pressure for antitrust regulation. Earlier this month, the US Federal Trade Commission filed an antitrust lawsuit against Facebook for its "anticompetitive conduct and unfair methods of competition." Over the summer, House Democrats published a 449-page report on Big Tech's monopolistic practices. The letter also comes in the context of rising geopolitical tensions.

Google employees have sent a letter to senior leadership demanding that the company reinstate and apologize to Timnit Gebru, a prominent Black researcher who said she was fired after criticizing the company's diversity efforts. Gebru's departure in early December has sparked outrage among Google staff and the industry at large. The letter – sent by Gebru's colleagues on Google's AI ethics team to the company's top management, including CEO Sundar Pichai – demands they offer Gebru a chance to return to the company at a higher position, in addition to making a public commitment to uphold research integrity and integrate racial literacy training. The letter also asks that management not to retaliate against the team for speaking out about Gebru's case. Bloomberg first reported news of the letter, which has been confirmed by the Guardian.