Human society is coordinated by mechanisms that control how prices are agreed, taxes are set, and electoral votes are tallied. The design of robust and effective mechanisms for human benefit is a core problem in the social, economic, and political sciences. Here, we discuss the recent application of modern tools from AI research, including deep neural networks trained with reinforcement learning (RL), to create more desirable mechanisms for people. We review the application of machine learning to design effective auctions, learn optimal tax policies, and discover redistribution policies that win the popular vote among human users. We discuss the challenge of accurately modeling human preferences and the problem of aligning a mechanism to the wishes of a potentially diverse group. We highlight the importance of ensuring that research into "deep mechanism design" is conducted safely and ethically.

Consider the process of collective decision-making, in which a group of individuals interactively select a preferred outcome from among a universe of alternatives. In this context, "representation" is the activity of making an individual's preferences present in the process via participation by a proxy agent – i.e. their "representative". To this end, learned models of human behavior have the potential to fill this role, with practical implications for multi-agent scenario studies and mechanism design. In this work, we investigate the possibility of training language agents to behave in the capacity of representatives of human agents, appropriately expressing the preferences of those individuals whom they stand for. First, we formalize the setting of collective decision-making – as the episodic process of interaction between a group of agents and a decision mechanism. On this basis, we then formalize the problem of digital representation – as the simulation of an agent's behavior to yield equivalent outcomes from the mechanism. Finally, we conduct an empirical case study in the setting of consensus-finding among diverse humans, and demonstrate the feasibility of fine-tuning large language models to act as digital representatives.

A canonical social dilemma arises when resources are allocated to people, who can either reciprocate with interest or keep the proceeds. The right resource allocation mechanisms can encourage levels of reciprocation that sustain the commons. Here, in an iterated multiplayer trust game, we use deep reinforcement learning (RL) to design a social planner that promotes sustainable contributions from human participants. We first trained neural networks to behave like human players, creating a stimulated economy that allows us to study the dynamics of receipt and reciprocation. We use RL to train a mechanism to maximise aggregate return to players. The RL mechanism discovers a redistributive policy that leads to a large but also more equal surplus. The mechanism outperforms baseline mechanisms by conditioning its generosity on available resources and temporarily sanctioning defectors. Examining the RL policy allows us to develop a similar but explainable mechanism that is more popular among players.

Effective approaches to encouraging group cooperation are still an open challenge. Here we apply recent advances in deep learning to structure networks of human participants playing a group cooperation game. We leverage deep reinforcement learning and simulation methods to train a ‘social planner’ capable of making recommendations to create or break connections between group members. The strategy that it develops succeeds at encouraging pro-sociality in networks of human participants ( N = 208 participants in 13 groups) playing for real monetary stakes. Under the social planner, groups finished the game with an average cooperation rate of 77.7%, compared with 42.8% in static networks ( N = 176 in 11 groups). In contrast to prior strategies that separate defectors from cooperators (tested here with N = 384 in 24 groups), the social planner learns to take a conciliatory approach to defectors, encouraging them to act pro-socially by moving them to small highly cooperative neighbourhoods.

This report introduces a new family of multimodal models, Gemini, that exhibit remarkable capabilities across image, audio, video, and text understanding. The Gemini family consists of Ultra, Pro, and Nano sizes, suitable for applications ranging from complex reasoning tasks to on-device memory-constrained use-cases. Evaluation on a broad range of benchmarks shows that our most-capable Gemini Ultra model advances the state of the art in 30 of 32 of these benchmarks - notably being the first model to achieve human-expert performance on the well-studied exam benchmark MMLU, and improving the state of the art in every one of the 20 multimodal benchmarks we examined. We believe that the new capabilities of the Gemini family in cross-modal reasoning and language understanding will enable a wide variety of use cases. We discuss our approach toward post-training and deploying Gemini models responsibly to users through services including Gemini, Gemini Advanced, Google AI Studio, and Cloud Vertex AI.

Building artificial intelligence (AI) that aligns with human values is an unsolved problem. Here we developed a human-in-the-loop research pipeline called Democratic AI, in which reinforcement learning is used to design a social mechanism that humans prefer by majority. A large group of humans played an online investment game that involved deciding whether to keep a monetary endowment or to share it with others for collective benefit. Shared revenue was returned to players under two different redistribution mechanisms, one designed by the AI and the other by humans. The AI discovered a mechanism that redressed initial wealth imbalance, sanctioned free riders and successfully won the majority vote. By optimizing for human preferences, Democratic AI offers a proof of concept for value-aligned policy innovation.

Recent work in large language modeling (LLMs) has used fine-tuning to align outputs with the preferences of a prototypical user. This work assumes that human preferences are static and homogeneous across individuals, so that aligning to a a single "generic" user will confer more general alignment. Here, we embrace the heterogeneity of human preferences to consider a different challenge: how might a machine help people with diverse views find agreement? We fine-tune a 70 billion parameter LLM to generate statements that maximize the expected approval for a group of people with potentially diverse opinions. Human participants provide written opinions on thousands of questions touching on moral and political issues (e.g., "should we raise taxes on the rich?"), and rate the LLM's generated candidate consensus statements for agreement and quality. A reward model is then trained to predict individual preferences, enabling it to quantify and rank consensus statements in terms of their appeal to the overall group, defined according to different aggregation (social welfare) functions. The model produces consensus statements that are preferred by human users over those from prompted LLMs (> 70%) and significantly outperforms a tight fine-tuned baseline that lacks the final ranking step. Further, our best model's consensus statements are preferred over the best human-generated opinions (> 65%). We find that when we silently constructed consensus statements from only a subset of group members, those who were excluded were more likely to dissent, revealing the sensitivity of the consensus to individual contributions. These results highlight the potential to use LLMs to help groups of humans align their values with one another.

Artificial learning agents are mediating a larger and larger number of interactions among humans, firms, and organizations, and the intersection between mechanism design and machine learning has been heavily investigated in recent years. However, mechanism design methods often make strong assumptions on how participants behave (e.g. rationality), on the kind of knowledge designers have access to a priori (e.g. access to strong baseline mechanisms), or on what the goal of the mechanism should be (e.g. total welfare). Here we introduce HCMD-zero, a general purpose method to construct mechanisms making none of these three assumptions. HCMD-zero learns to mediate interactions among participants and adjusts the mechanism parameters to make itself more likely to be preferred by participants. It does so by remaining engaged in an electoral contest with copies of itself, thereby accessing direct feedback from participants. We test our method on a stylized resource allocation game that highlights the tension between productivity, equality and the temptation to free ride. HCMD-zero produces a mechanism that is preferred by human participants over a strong baseline, it does so automatically, without requiring prior knowledge, and using human behavioral trajectories sparingly and effectively. Our analysis shows HCMD-zero consistently makes the mechanism policy more and more likely to be preferred by human participants over the course of training, and that it results in a mechanism with an interpretable and intuitive policy.

From social networks to traffic routing, artificial learning agents are playing a central role in modern institutions. We must therefore understand how to leverage these systems to foster outcomes and behaviors that align with our own values and aspirations. While multiagent learning has received considerable attention in recent years, artificial agents have been primarily evaluated when interacting with fixed, non-learning co-players. While this evaluation scheme has merit, it fails to capture the dynamics faced by institutions that must deal with adaptive and continually learning constituents. Here we address this limitation, and construct agents ("mechanisms") that perform well when evaluated over the learning trajectory of their adaptive co-players ("participants"). The algorithm we propose consists of two nested learning loops: an inner loop where participants learn to best respond to fixed mechanisms; and an outer loop where the mechanism agent updates its policy based on experience. We report the performance of our mechanism agents when paired with both artificial learning agents and humans as co-players. Our results show that our mechanisms are able to shepherd the participants strategies towards favorable outcomes, indicating a path for modern institutions to effectively and automatically influence the strategies and behaviors of their constituents.

In standard neural networks the amount of computation used grows with the size of the inputs, but not with the complexity of the problem being learnt. To overcome this limitation we introduce PonderNet, a new algorithm that learns to adapt the amount of computation based on the complexity of the problem at hand. PonderNet learns end-to-end the number of computational steps to achieve an effective compromise between training prediction accuracy, computational cost and generalization. On a complex synthetic problem, PonderNet dramatically improves performance over previous adaptive computation methods and additionally succeeds at extrapolation tests where traditional neural networks fail. Also, our method matched the current state of the art results on a real world question and answering dataset, but using less compute. Finally, PonderNet reached state of the art results on a complex task designed to test the reasoning capabilities of neural networks.1

When making economic choices, such as those between goods or gambles, humans act as if their internal representation of the value and probability of a prospect is distorted away from its true value. These distortions give rise to decisions which apparently fail to maximise reward, and preferences that reverse without reason. Why would humans have evolved to encode value and probability in a distorted fashion, in the face of selective pressure for reward-maximising choices? Here, we show that under the simple assumption that humans make decisions with finite computational precision – in other words, that decisions are irreducibly corrupted by noise – the distortions of value and probability displayed by humans are approximately optimal in that they maximise reward and minimise uncertainty. In two empirical studies, we manipulate factors that change the reward-maximising form of distortion, and find that in each case, humans adapt optimally to the manipulation. This work suggests an answer to the longstanding question of why humans make “irrational” economic choices.

Humans and other animals make decisions in order to satisfy their goals. However, it remains unknown how neural circuits compute which of multiple possible goals should be pursued (e.g., when balancing hunger and thirst) and how to combine these signals with estimates of available reward alternatives. Here, humans undergoing fMRI accumulated two distinct assets over a sequence of trials. Financial outcomes depended on the minimum cumulate of either asset, creating a need to maintain "value equilibrium" by redressing any imbalance among the assets. Blood-oxygen-level-dependent (BOLD) signals in the rostral anterior cingulate cortex (rACC) tracked the level of imbalance among goals, whereas the ventromedial prefrontal cortex (vmPFC) signaled the level of redress incurred by a choice rather than the overall amount received. These results suggest that a network of medial frontal brain regions compute a value signal that maintains value equilibrium among internal goals.

Humans can learn to perform multiple tasks in succession over the lifespan (“continual” learning), whereas current machine learning systems fail. Here, we investigated the cognitive mechanisms that permit successful continual learning in humans. Unlike neural networks, humans that were trained on temporally autocorrelated task objectives (focussed training) learned to perform new tasks more effectively, and performed better on a later test involving randomly interleaved tasks. Analysis of error patterns suggested that focussed learning permitted the formation of factorised task representations that were protected from mutual interference. Furthermore, individuals with a strong prior tendency to represent the task space in a factorised manner enjoyed greater benefit of focussed over interleaved training. Building artificial agents that learn to factorise tasks appropriately may be a promising route to solving continual task performance in machine learning. Significance Statement Humans learn to perform many different tasks over the lifespan, such as speaking both French and Spanish. The brain has to represent task information without mutual interference. In machine learning, this "continual learning" is a major unsolved challenge. Here, we studied the patterns of errors made by humans and state-of-the-art deep networks whilst they learned new tasks from scratch and without instruction. Humans, but not machines, seem to benefit from training regimes that focussed on one task at a time, especially when they had a prior bias to represent stimuli in a way that facilitated task separation. Machines trained to exhibit the same prior bias suffered less interference between tasks, suggesting new avenues for solving continual learning in artificial systems.

Humans and other animals make decisions in order to satisfy their goals. However, it remains unknown how neural circuits compute which of multiple possible goals should be pursued (e.g. when balancing hunger and thirst) and combine these signals with estimates of available reward alternatives. Here, humans undergoing functional magnetic resonance imaging (fMRI) accumulated two distinct assets over a sequence of trials. Financial outcomes depended on the minimum cumulate of either asset, creating a need to maintain “value homeostasis” by redressing any imbalance among the assets. BOLD signals in the dorsal anterior cingulate cortex (dACC) tracked the level of homeostatic imbalance among goals, whereas the ventromedial prefrontal cortex (vmPFC) signalled the level of homeostatic redress incurred by a choice, rather than the overall amount received. These results suggest that a network of medial frontal brain regions compute a value signal that maintains homeostatic balance among internal goals.

When making decisions, humans are often distracted by irrelevant information. Distraction has a different impact on perceptual, cognitive, and value-guided choices, giving rise to well-described behavioral phenomena such as the tilt illusion, conflict adaptation, or economic decoy effects. However, a single, unified model that can account for all these phenomena has yet to emerge. Here, we offer one such account, based on adaptive gain control, and additionally show that it successfully predicts a range of counterintuitive new behavioral phenomena on variants of a classic cognitive paradigm, the Eriksen flanker task. We also report that blood oxygen level-dependent signals in a dorsal network prominently including the anterior cingulate cortex index a gain-modulated decision variable predicted by the model. This work unifies the study of distraction across perceptual, cognitive, and economic domains.

Humans can learn to perform multiple tasks in succession over the lifespan ("continual" learning), whereas current machine learning systems fail. Here, we investigated the cognitive mechanisms that permit successful continual learning in humans and harnessed our behavioral findings for neural network design. Humans categorized naturalistic images of trees according to one of two orthogonal task rules that were learned by trial and error. Training regimes that focused on individual rules for prolonged periods (blocked training) improved human performance on a later test involving randomly interleaved rules, compared with control regimes that trained in an interleaved fashion. Analysis of human error patterns suggested that blocked training encouraged humans to form "factorized" representation that optimally segregated the tasks, especially for those individuals with a strong prior bias to represent the stimulus space in a well-structured way. By contrast, standard supervised deep neural networks trained on the same tasks suffered catastrophic forgetting under blocked training, due to representational interference in the deeper layers. However, augmenting deep networks with an unsupervised generative model that allowed it to first learn a good embedding of the stimulus space (similar to that observed in humans) reduced catastrophic forgetting under blocked training. Building artificial agents that first learn a model of the world may be one promising route to solving continual task performance in artificial intelligence research.

Humans and other animals accumulate resources, or wealth, by making successive risky decisions. If and how risk attitudes vary with wealth remains an open question. Here humans accumulated reward by accepting or rejecting successive monetary gambles within arbitrarily defined temporal contexts. Risk preferences changed substantially toward risk aversion as reward accumulated within a context, and blood oxygen level dependent (BOLD) signals in the ventromedial prefrontal cortex (PFC) tracked the latent growth of cumulative economic outcomes. Risky behavior was captured by a computational model in which reward prompts an adaptive update to the function that links utilities to choices. These findings can be understood if humans have evolved economic decision policies that fail to maximize overall expected value but reduce variance in cumulative outcomes, thereby ensuring that resources remain above a critical survival threshold.

When making decisions, humans are often distracted by irrelevant information. Distraction has different impact on perceptual, cognitive and value-guided choices, giving rise to well-described behavioural phenomena such as the tilt illusion, conflict adaptation, or economic decoy effects. However, a single, unified model that can account for all these phenomena has yet to emerge. Here, we offer one such account, based on adaptive gain control, and additionally show that it successfully predicts a range of counterintuitive new behavioural phenomena on variants of a classic cognitive paradigm, the Eriksen flanker task. We further report that BOLD signals in the dorsal anterior cingulate cortex (dACC), rather than reflecting the cost of resolving conflict among responses, index a gain-modulated decision variable predicted by the model. This work unifies the study of distraction across perceptual, cognitive and economic domains, and offers new insights into the puzzling pattern of neural responses typically observed in the dACC.