Search

I formalize the problem of care in the mathematical language of sequential decision-making. Drawing upon insights from developmental psychology, robotics, and computational cognitive modeling, I conceptualize care as a dynamic interplay between the caregiver ('one-caring') and the care recipient ('cared-for'). Caring actions maximize the utility of the cared-for at a future point when they are required to act autonomously. Since this quantity cannot be directly optimized, the focus is on enabling increasing levels of autonomy through environmental shaping, risk reduction, and safe exploration. I distinguish caregiving from helping and teaching by care's focus on exploration and autonomy that increase capacity over time. In the context of elderly care, the emphasis shifts towards preserving rather than enhancing capacity. Finally, I consider the role of caregiving in the development of moral values and the possibility of artificially intelligent agents that might someday care for us.

Even simple, ambiguous signals can have a rich interpretation when viewed in the context of an interaction in a shared environment. We create a model called Paternalistic Communication by combining an existing model of overloaded language -- Rational Speech Acts (RSA) -- with a full agent model of Theory of Mind (ToM). This integration allows signals to be processed in conjunction with common ground in a principled manner dependent on task-dependent action utilities. This modeling perspective treats communication as a way to coordinate diverging perspectives in a cooperative setting. Under Paternalistic Communication, a speaker decides what to say by predicting their partner’s reaction based on the information in common ground and then evaluates those reactions using their own mind which may contain additional information. We demonstrate the flexibility and performance of Paternalistic Communication in a case study with ambiguous signaling through a set of simulations.

Humans navigate daily decision-making by flexibly choosing appropriate approximations of what ought to be done. Which mental algorithms do people use, and when? We use behavioural experiments and modelling to investigate three computational principles known to influence decision making: noisy utility maximization, discounting, and the probability warping principle of Prospect Theory.While these principles have been shown to separately influence human behaviour in simple laboratory tasks, such as bandits and gambles, we evaluate their combined use in the context of a naturalistic spatial search that required sequential decision-making. We found that while aggregate human behaviour can be reasonably well explained by an optimal planner with noisy utility maximization, individual-level behaviour exhibits consistent irregularities, that deviate from expected utility theory.We show that model-based prediction of individual-level behaviours in our experiment is significantly improved by combining the three computational principles, and benefits particularly strongly from probability warping. Furthermore, our results suggest that probability warping may be a common factor of human decision making, that generalizes beyond the gambles explored in Prospect Theory, to natural human behaviours such as spatial search and navigation.

Collaboration requires agents to coordinate their behavior onthe fly, sometimes cooperating to solve a single task togetherand other times dividing it up into sub-tasks to work on in par-allel. Underlying the human ability to collaborate is theory-of-mind, the ability to infer the hidden mental states that driveothers to act. Here, we develop Bayesian Delegation, a decen-tralized multi-agent learning mechanism with these abilities.Bayesian Delegation enables agents to rapidly infer the hid-den intentions of others by inverse planning. These inferencesenable agents to flexibly decide in the absence of communi-cation when to cooperate on the same sub-task and when towork on different sub-tasks in parallel. We test this model ina suite of multi-agent Markov decision processes inspired bycooking problems. To succeed, agents must coordinate boththeir high-level plans (e.g., what sub-task they should work on)and their low-level actions (e.g., avoiding collisions). BayesianDelegation bridges these two levels and rapidly aligns agents’beliefs about who should work on what. Finally, we testedBayesian Delegation in a behavioral experiment where partici-pants made sub-task inferences from sparse observations of co-operative behavior. Bayesian Delegation outperformed heuris-tic models and was closely aligned with human judgments.

Cumulative culture depends on the fidelity of learning be-tween successive generations, and the robustness with whichthe lessons of one generation apply to the problems of the next.How do humans accomplish these twin goals? We formalizesocial learning as a kind of program induction, and provide anexperimental test of a key prediction. To do this, we exploit akey fact: When humans learn from others, in addition to ob-serving inputs and outputs we often observe the process thatled to that output. For instance, when preparing a meal, wedon’t just observe a pile of vegetables and then a ratatouille.Instead, we observe a causal process that transforms those in-gredients into a finished food. Here, we use probabilistic pro-grams to represent causal processes and show that the observa-tion of an execution trace speeds up program induction, evenwhen learning from only a single example. This model pre-dicts that the inferences and behavior of people will be struc-tured by these execution traces. In two behavioral experiments,we show that human judgments and behavior are affected bythe execution trace in the systematic ways predicted by our for-mal model. These findings shed light on the mechanisms thatunderlie high fidelity social learning in humans, and unify therole of emulation and imitation in social learning.

Communication is highly overloaded. Despite this, even young children are good at leveraging context to understandambiguous signals. We propose a computational shared agency account of signaling that we call the Imagined We (IW)framework. We leverage Bayesian Theory of Mind to provide mechanisms for rational action planning and inverse actioninterpretation. In order to expand this framework for communication, we first treat signals as rational actions. We thenincorporate our rich understanding of intuitive utilities to constrain the scope of affordable actions. Finally, we treatcommunication as a cooperative act, subject to constraints of maximizing a shared utility. We implement this modelin two completely different behavioral psychology works to demonstrate the generality of the IW under different typesof uncertainty in cooperative communication. Additionally, we demonstrate that the IW outperforms multiple baselinemodels in a novel task across a series of simulation conditions.

We describe a cognitive mechanism of moral judgment, universalization, that has received little attention up to now. Underuniversalization, an action’s moral permissibility is determined by calculating what the outcome would be if all people whoare similarly situated to the actor also acted in that way. This mechanism is particularly well-suited to capture our moraljudgments of free-rider cases, where one person doing the action increases utility but many people doing it decreasesutility. Universalization fits into an agreement-based (contractualist) theory of moral cognition, and explains properties ofour moral judgments that an outcome-based or rule-based approach cannot. We show patterns of universalization reasoningin young children as well as adults.

We present a generative model of how observers think about the emotions experienced by players in a socially-chargedgame: a public, high-stakes, one-shot Prisoner’s Dilemma. The model extends inverse planning frameworks to captureobservers’ judgments about players’ reactions to hypothetical events. Observers attribute different beliefs and values toplayers based on what decisions the players make. We model how observers’ noisy inferences of players’ mental contentsbias emotion predictions. Incorporation of non-monetary features into forward planning enables us to model emotions thatreflect complex social concerns (e.g. Embarrassment depends on how much players think others will infer that they tried totake advantage of their opponents). In addition to matching the intensities of twenty attributed emotions, the model reflectshow observers’ emotion judgments covary within single stimuli, indicating that the model captures important aspects ofthe generative process underlying humans’ emotion attributions in this game.

Cumulative culture ultimately depends on the fidelity of learning between successive generations. When humans learnfrom others in addition to observing inputs and outputs we often observe the process which led to that output. Forinstance, when preparing a meal we don’t just observe a pile of vegetables and then a ratatouille. Instead, we observe acausal process by which those ingredients are transformed. Here we use programs to represent a cultural process and showthat the observation of an execution trace speeds up program induction even when learning from only a single example.This mechanism could account for (1) the high fidelity of social learning which leads to cumulative culture in humans(2) unify the role of emulation and imitation in social learning and (3) account for aspects of moral learning such asritualization.

Discrete choice models (e.g. logistic regression) are popular models in the economics literature that describe choices between twoor more discrete alternatives. These models have been successfully used to model value-based decisions, e.g. decisions in moraldilemmas, although temporal components of a decision, such as reaction times and changes of mind are not included. In cognitivesciences, another class of decision models, namely sequential-sampling models, has gained popularity in modelling choice accuracy,reaction time and decision uncertainty (e.g. confidence judgments). Here, we model decisions in moral dilemmas using a variant ofa hierarchical drift-diffusion model, factor drift diffusion, that combines the value-based approach with that of evidence accumulationmechanism by sequential-sampling. Specifically, we model the evidence accumulation process as resulting from a subjective weightingof abstract moral dimensions (factors). We train our model on a data set of 6500 moral decisions by 500 respondents on a popularweb platform (MoralMachine.mit.edu) and separately infer different sources of uncertainty in moral decisions. We show that the modelsuccessfully predicts reaction times and choices in moral dilemmas, while also leading to unexpected results

We propose a formal model of humans’ intuitive theories of others’ emotions. From a single choice in a social interaction(e.g. the choice to cooperate in a Prisoner’s Dilemma game), human observers can infer a player’s complex values, such asprosocial preferences and reputational concerns. When the player then experiences a new situation (the game’s outcome),observers infer the player’s reaction to the event based on the mental state likely to have produced the player’s action. Herewe capture this process by inverting a richly structured generative model of social gameplay, including social equity andreputational dimensions, and translate players’ subjective motivations, expectations, and prediction errors into forwardpredictions of the emotional experiences of the players. Our model infers players’ values and expectations, generatespatterns of play that match observers’ intuitions, and supports formally generated emotion predictions with substantiallyextended breadth and nuance.

Perseverance, above and beyond IQ, predicts academic outcomes in school age children, however, little is knownabout what factors affect persistence in early childhood. Here, we propose a formal Bayesian model of how children mightlearn how to calibrate effort from observing adult models and then explore this idea behaviorally across two experiments inchildren and infants. Results from Experiment 1 show that preschoolers persist more after watching an adult persist, but onlyif the adult is successful at reaching their goal. Experiment 2 and a pre-registered replication extend these findings, showingthat even infants use adult models to modulate their persistence, and can generalize this inference to novel situations. Theseresults suggest that both preschoolers and infants are sensitive to adult persistence and use it to calibrate their own effort infar-reaching ways.

Successful and repeated cooperation requires fairly sharingthe spoils of joint endeavors. Fair distribution is often doneaccording to preferences for equitable outcomes even thoughstrictly equitable outcomes can lead to inefficient waste. In ad-dition to preferences about the outcome itself, decision makersare also sensitive to the attributions others might make aboutthem as a result of their choice. We develop a novel mathemat-ical model where decision makers turn their capacity to inferlatent desires and beliefs from the behavior of others (theory-of-mind) towards themselves, anticipating the judgments oth-ers will make about them. Using this model we can construct apreference to be seen as impartial and integrate it with prefer-ences for equitable and efficient outcomes. We test this modelin two studies where the anticipated attribution of impartialityis ambiguous: when one agent is more deserving than the otherand when unbiased procedures for distribution are made avail-able. This model explains both participants’ judgments aboutthe partiality of others and their hypothetical decisions. Ourmodel argues that people avoid inequity not only because theyfind it inherently undesirable, they also want to avoid beingjudged as partial.

People often use norms to coordinate behavior andaccomplish shared goals. But how do people learn andrepresent norms? Here, we formalize the process by whichcollaborating individuals (1) reason about group plans duringinteraction, and (2) use task features to abstractly representnorms. In Experiment 1, we test the assumptions of our modelin a gridworld that requires coordination and contrast it with a“best response” model. In Experiment 2, we use our model totest whether group members’ joint planning relies more onstate features independent of other agents (landmark-basedfeatures) or state features determined by the configuration ofagents (agent-relative features).

Successfully navigating the social world requires reasoningabout both high-level strategic goals, such as whether to co-operate or compete, as well as the low-level actions neededto achieve those goals. While previous work in experimentalgame theory has examined the former and work on multi-agentsystems has examined the later, there has been little work in-vestigating behavior in environments that require simultaneousplanning and inference across both levels. We develop a hierar-chical model of social agency that infers the intentions of otheragents, strategically decides whether to cooperate or competewith them, and then executes either a cooperative or competi-tive planning program. Learning occurs across both high-levelstrategic decisions and low-level actions leading to the emer-gence of social norms. We test predictions of this model inmulti-agent behavioral experiments using rich video-game likeenvironments. By grounding strategic behavior in a formalmodel of planning, we develop abstract notions of both co-operation and competition and shed light on the computationalnature of joint intentionality.

Many social judgments hinge on assigning responsibility to individualsfor their role in a group’s success or failure. Often thegroup’s success depends on every team member acting in a rationalway. When someone does not conform to what othersexpect of them, cooperation breaks down. We present a computationalmodel of responsibility judgments for individualsin a cooperative setting. We test the model in two behavioralexperiments where participants were asked to evaluate agentsacting in a cooperative, one-shot game. In Experiment 1, weshow that participants’ action predictions are consistent with arecursive reasoning model. In Experiment 2, we show that people’sassignments of blame are influenced by both an agent’spresumed rationality, or adherence to an expected policy, aswell as the pivotality of the agent’s actions, or how close thesituation was to one in which the action would have made adifference to the outcome.

The actions of a rational agent reveal information about itsmental states. These inferred mental states, particularly theagent’s intentions, play an important role in the evaluation ofmoral permissibility. While previous computational modelshave shown that beliefs and desires can be inferred from behaviorunder the assumption of rational action they have criticallylacked a third mental state, intentions. In this work, wedevelop a novel formalism for intentions and show how theycan be inferred as counterfactual contrasts over influence diagrams.This model is used to quantitatively explain judgmentsabout intention and moral permissibility in classic and noveltrolley problems.

The empirical study of virtue is plagued by imprecise definitions and assessment. Here we propose a three-stage, data-driven ('bottom-up') method to differentiate lay perceptions of virtues. Employing two virtues – generosity (as cooperation) and fairness (as impartiality) – as a case study, we present findings utilizing data from three studies (total N = 2,667). First, natural language processing of free-response data indicated that participants used different 'topics' (i.e. clusters of words) to describe behaviours representing generosity (topics: 'charity' and 'kindness') and fairness ('equality'). Second, participants in a survey experiment rated behaviours expressing generosity and fairness differently across 6 out of 9 underlying features measured. Third, participants perceive that actors in vignette-based experiments engaging in behaviours expressing generosity versus fairness were motivated differently on 5 out of 6 motivations measured. Our findings support the distinction of the virtues of generosity (as cooperation) and fairness (as impartiality) and indicate the utility of our bottom-up method for assessing and distinguishing virtues. [ABSTRACT FROM AUTHOR]