Intelligent decision support in medical triage: are people robust to biased advice?

BACKGROUND: Intelligent artificial agents ('agents') have emerged in various domains of human society (healthcare, legal, social). Since using intelligent agents can lead to biases, a common proposed solution is to keep the human in the loop. Will this be enough to ensure unbiased decision making? METHODS: To address this question, an experimental testbed was developed in which a human participant and an agent collaboratively conduct triage on patients during a pandemic crisis. The agent uses data to support the human by providing advice and extra information about the patients. In one condition, the agent provided sound advice; the agent in the other condition gave biased advice. The research question was whether participants neutralized bias from the biased artificial agent. RESULTS: Although it was an exploratory study, the data suggest that human participants may not be sufficiently in control to correct the agent's bias. CONCLUSIONS: This research shows how important it is to design and test for human control in concrete human-machine collaboration contexts. It suggests that insufficient human control can potentially result in people being unable to detect biases in machines and thus unable to prevent machine biases from affecting decisions.

Identifying Interaction Patterns of Tangible Co-Adaptations in Human-Robot Team Behaviors.

As robots become more ubiquitous, they will increasingly need to behave as our team partners and smoothly adapt to the (adaptive) human team behaviors to establish successful patterns of collaboration over time. A substantial amount of adaptations present themselves through subtle and unconscious interactions, which are difficult to observe. Our research aims to bring about awareness of co-adaptation that enables team learning. This paper presents an experimental paradigm that uses a physical human-robot collaborative task environment to explore emergent human-robot co-adaptions and derive the interaction patterns (i.e., the targeted awareness of co-adaptation). The paradigm provides a tangible human-robot interaction (i.e., a leash) that facilitates the expression of unconscious adaptations, such as "leading" (e.g., pulling the leash) and "following" (e.g., letting go of the leash) in a search-and-navigation task. The task was executed by 18 participants, after which we systematically annotated videos of their behavior. We discovered that their interactions could be described by four types of adaptive interactions: stable situations, sudden adaptations, gradual adaptations and active negotiations. From these types of interactions we have created a language of interaction patterns that can be used to describe tacit co-adaptation in human-robot collaborative contexts. This language can be used to enable communication between collaborating humans and robots in future studies, to let them share what they learned and support them in becoming aware of their implicit adaptations.

Becoming Team Members: Identifying Interaction Patterns of Mutual Adaptation for Human-Robot Co-Learning.

Becoming a well-functioning team requires continuous collaborative learning by all team members. This is called co-learning, conceptualized in this paper as comprising two alternating iterative stages: partners adapting their behavior to the task and to each other (co-adaptation), and partners sustaining successful behavior through communication. This paper focuses on the first stage in human-robot teams, aiming at a method for the identification of recurring behaviors that indicate co-learning. Studying this requires a task context that allows for behavioral adaptation to emerge from the interactions between human and robot. We address the requirements for conducting research into co-adaptation by a human-robot team, and designed a simplified computer simulation of an urban search and rescue task accordingly. A human participant and a virtual robot were instructed to discover how to collaboratively free victims from the rubbles of an earthquake. The virtual robot was designed to be able to real-time learn which actions best contributed to good team performance. The interactions between human participants and robots were recorded. The observations revealed patterns of interaction used by human and robot in order to adapt their behavior to the task and to one another. Results therefore show that our task environment enables us to study co-learning, and suggest that more participant adaptation improved robot learning and thus team level learning. The identified interaction patterns can emerge in similar task contexts, forming a first description and analysis method for co-learning. Moreover, the identification of interaction patterns support awareness among team members, providing the foundation for human-robot communication about the co-adaptation (i.e., the second stage of co-learning). Future research will focus on these human-robot communication processes for co-learning.

Moral Decision Making in Human-Agent Teams: Human Control and the Role of Explanations.

With the progress of Artificial Intelligence, intelligent agents are increasingly being deployed in tasks for which ethical guidelines and moral values apply. As artificial agents do not have a legal position, humans should be held accountable if actions do not comply, implying humans need to exercise control. This is often labeled as Meaningful Human Control (MHC). In this paper, achieving MHC is addressed as a design problem, defining the collaboration between humans and agents. We propose three possible team designs (Team Design Patterns), varying in the level of autonomy on the agent's part. The team designs include explanations given by the agent to clarify its reasoning and decision-making. The designs were implemented in a simulation of a medical triage task, to be executed by a domain expert and an artificial agent. The triage task simulates making decisions under time pressure, with too few resources available to comply with all medical guidelines all the time, hence involving moral choices. Domain experts (i.e., health care professionals) participated in the present study. One goal was to assess the ecological relevance of the simulation. Secondly, to explore the control that the human has over the agent to warrant moral compliant behavior in each proposed team design. Thirdly, to evaluate the role of agent explanations on the human's understanding in the agent's reasoning. Results showed that the experts overall found the task a believable simulation of what might occur in reality. Domain experts experienced control over the team's moral compliance when consequences were quickly noticeable. When instead the consequences emerged much later, the experts experienced less control and felt less responsible. Possibly due to the experienced time pressure implemented in the task or over trust in the agent, the experts did not use explanations much during the task; when asked afterwards they however considered these to be useful. It is concluded that a team design should emphasize and support the human to develop a sense of responsibility for the agent's behavior and for the team's decisions. The design should include explanations that fit with the assigned team roles as well as the human cognitive state.

Long-term care use and socio-economic status in Belgium: a survival analysis using health care insurance data.

BACKGROUND: The small but growing literature on socio-economic inequality in morbidity among older persons suggests that social inequalities in health persist into old age. A largely separate body of literature looks at the predictors of long-term care use, in particular of institutional care. Various measures of socio-economic status are often included as control variables in these studies. Review articles generally conclude that the evidence for such variables being a predictor for institutionalization is "inconclusive". In this paper we look at the association among older persons in Belgium between one particular measure of socio-economic status - preferential status in public health care insurance - and first use of home long-term care and residential care. Preferential status entitles persons to higher reimbursement rates for health care from the public health care insurance system and is conditional on low income. We also study whether preferential status is related to the onset of five important chronic conditions and the time of death. METHODS: We use survival analysis; the source of the data is a large administrative panel of a sample representative for all older persons in Belgium (1,268,740 quarterly observations for 69,562 individuals). RESULTS: We find a strong association between preferential status and the likelihood of home care use, but for residential care it is small for men and non-existent for women. We also find that preferential status is significantly related to the chance of getting two out five chronic conditions - COPD and diabetes, but not dementia, hip fracture and Parkinson's disease - and to the probability of dying (not for women). For home care use and death, the association with preferential status declines with increasing age from age 65 onwards, such that it is near zero for those aged around 90 and older. CONCLUSION: We find clear associations between an indicator of low income and home care use, some chronic conditions and death. The associations are stronger among men than among women. We also find that the association declines with age for home care use and death, which might be explained by selective survival.

Transitions in formal and informal care utilisation amongst older Europeans: the impact of national contexts.

The objective of this study was to explore how long-term care systems, and in particular the incorporation of needs-based entitlements to care services or benefits, influence formal and informal care utilisation dynamics. We used the Survey of Health, Ageing and Retirement in Europe (SHARE) wave 1 and 2 data, restricting the sample to persons 65+ from 9 European countries (N = 6,293). The effects of changes in health and household composition on formal and informal care transitions were estimated using logistic regression, allowing these effects to vary across countries. The results indicated that, in all countries, formal and informal care were more often complements than substitutes. The likelihood of becoming a formal or informal care user varied significantly between countries. In the Scandinavian countries and in several continental European countries with needs-based entitlements, the transition to formal care was strongly related to informal support being or becoming unavailable. We found little evidence of country differences in the effect of health variables on the transition to formal care. The analysis suggested that, whilst rates of formal care utilisation continue to differ considerably between European countries, formal care allocation practices are not very dissimilar across Northern and continental European welfare states, as we found evidence for all countries of targeting of older persons living alone and of the most care-dependent older people.

Like mother, like daughter? Mother's history of cervical cancer screening and daughter's Human Papillomavirus vaccine uptake in Flanders (Belgium).

OBJECTIVE: We investigated whether and to what extent the uptake of the Human Papillomavirus (HPV) vaccine by girls aged 12-18 was related to the cervical cancer screening history of age-appropriate older female household members (assumed to be their mothers) in Flanders (Belgium). METHODS: We studied administrative records on 127,854 female members of the National Alliance of Christian Mutualities, which is the largest health insurance fund in Flanders. Reimbursement data for HPV vaccination of girls for the period 2007-2009 were linked with reimbursement data for cervical cancer screening of their mothers in the three preceding years. A multilevel logit model was used to study associations between both preventive behaviors. In the model we controlled for both the girl's and the mother's age, the province of residence and the socio-economic background of the family. RESULTS: A clear association between a mother's history of participation in cervical cancer screening and her daughter's HPV vaccination initiation was found. The conditional odds of HPV vaccination initiation were more than 4 times higher for girls whose mother had one Pap test than for girls whose mother had none (odds ratio [OR]=4.5; 95% confidence interval [CI]=3.5-5.9). For girls whose mother had three or more Pap tests, the conditional odds were 16 times higher than for girls whose mother did not have any pap tests ([OR]=16.0; 95% [CI]=12.1-21.2). The effect of screening (having received 1 pap smear as compared to none) was larger for girls living in neighborhoods with the lowest median income ([OR]=6.0, 95% [CI]=3.6-10.1). CONCLUSION: In a situation where both cervical cancer screening and HPV vaccination are opportunistic, we found evidence that these preventive behaviors cluster within families.

The effects of critical thinking instruction on training complex decision making.

OBJECTIVE: Two field studies assessed the effects of critical thinking instruction on training and transfer of a complex decision-making skill. BACKGROUND: Critical thinking instruction is based on studies of how experienced decision makers approach complex problems. METHOD: Participants conducted scenario-based exercises in both simplified (Study I) and high-fidelity (Study 2) training environments. In both studies, half of the participants received instruction in critical thinking. The other half conducted the same exercises but without critical thinking instruction. After the training, test scenarios were administered to both groups. RESULTS: The first study showed that critical thinking instruction enhanced decision outcomes during both training and the test. In the second study, critical thinking instruction benefited both decision outcomes and processes, specifically on the transfer to untrained problems. CONCLUSION: The results suggest that critical thinking instruction improves decision strategy and enhances understanding of the general principles of the domain. APPLICATION: The results of this study warrant the implementation of critical thinking instruction in training programs for professional decision makers that have to operate in complex and highly interactive, dynamic environments.