Artificial Intelligence (AI) systems have significantly grown in importance in recent years, handling tasks and making decisions that were once exclusive to humans. Substantial progress in AI research has yielded highly accurate systems that can even outperform human capabilities. However, it is premature to rely on autonomous intelligent systems blindly; human-AI collaboration remains critical to achieving optimal outcomes in diverse domains. Examples of successful collaboration include driverless systems for car travel, automated diagnostic tools in healthcare, and scoring systems in the financial industry.
Image credit: Blue Planet Studio/Shutterstock
Human-AI Collaboration
In recent years, AI systems have taken on tasks and decision-making once limited to humans, demonstrating tremendous progress in AI research. While AI systems can surpass human capabilities in certain areas, blind reliance on autonomous intelligent systems is premature. Human-AI collaboration remains essential for achieving optimal outcomes in various domains. Examples of such collaboration include driverless systems for car travel, automated diagnostic tools in healthcare, and scoring systems in the financial industry.
Researchers are actively investigating Human-AI collaboration, aiming to develop mental models of AI systems that enable users to understand their error boundaries. The goal is to achieve optimal team performance by fostering calibrated trust in the AI system. However, users often lack awareness of the model's error boundary, leading to uncalibrated trust and a mismatch between AI trust and the AI's actual trustworthiness. This discrepancy can result in excessive reliance on the AI system or inadequate reliance, potentially compromising decision-making.
To address these challenges, research on Human-AI collaboration explores the dynamic between AI systems and humans working as a team. Collaborative efforts between Human-Computer Interaction (HCI) and Computer-supported cooperative work (CSCW) researchers and AI experts can drive a human-in-the-loop AI initiative. This concept of Human-AI collaboration dates back to J.C.R. Licklider's notion of symbiotic computing in 1960. Today, AI-powered systems, such as AI-based clinical decision support systems, customer service chatbots, and automated machine learning, are designed to complement and enhance the work of professionals in various fields.
Collaborative Intelligence
Organizations can harness the power of collaborative intelligence by understanding how humans can effectively augment machines and vice versa.
Humans Assisting Machines: Human trainers play a crucial role in teaching machine-learning algorithms specific tasks and instilling personalities in AI assistants. Explainers are vital for making AI decisions transparent, especially in evidence-based industries. Sustainers ensure AI systems function safely and responsibly, with roles ranging from safety engineers to ethics managers and data compliance officers. This collaboration creates new job opportunities and enhances the positive impact of AI in various industries.
Machines Assisting Humans: AI systems are enhancing human capabilities in three ways: amplifying cognitive strengths, enabling better interactions with customers and employees, and extending physical capabilities through embodied intelligence. For instance, AI like Dreamcatcher can enhance creativity by generating thousands of design ideas, while AI assistants like Aida enable scalable and efficient customer interactions. Moreover, AI-powered robots are collaborating with humans in manufacturing processes.
To leverage this collaboration effectively, businesses must redesign their operations, co-create solutions with AI, and scale and sustain the proposed improvements. AI also assists in decision-making and enables personalized customer experiences, revolutionizing industries such as marketing and hospitality.
Trust in collaboration: Trust plays a central role in human trust towards AI and AI's trust in humans. Trust dynamics are complex, vary over time, and are conditional based on the system's predictability and dependability. Expert users of autonomous systems exhibit deliberate trust, considering the conditions for trusting the system.
In Human-AI collaborations, trust calibration works bi-directionally, with humans and autonomous agents verifying each other's declarations and instructions. This bi-directional trust is crucial for effective team performance, preventing unintended consequences, and ensuring optimal collaboration between humans and machines.
Applications of Human-AI Collaboration
Healthcare: AI systems are increasingly valuable in medical decision-making, supporting Medical Doctors (MDs) across various domains. The collaboration between humans and AI, termed "hybrid intelligence," promises superior outcomes. However, challenges like over-reliance, under-reliance, and opacity of judgments' reliability must be addressed to create an effective human-AI team. Research in the clinical domain explores human-AI collaboration, with AI-assisted colonoscopy as a case study, showcasing the complementary roles of endoscopists and AI.
Education: Human-AI collaboration in education requires a comprehensive approach. AI in education is evolving from assessing student knowledge to diagnosing various learner characteristics and progress features. Self-regulated learning (SRL), emotion, motivation, engagement, and collaboration are emerging research areas. AI-supported learning technologies raise concerns about learners' loss of control, but hybrid intelligence can combine learner and AI regulation.
AI detects SRL through data streams such as clickstreams, eye tracking, and physiological data, leading to an accurate diagnosis. Diagnostics inform pedagogical actions, such as personalized scaffolds and dashboards. Developing AI solutions requires co-creation involving researchers, learners, teachers, and developers to advance the augmentation perspective and improve AI integration in education.
Finance: AI will revolutionize banking in the next century, making it more personalized, customer-centric, and efficient. Banks and FinTech startups are heavily investing in AI, using robots and chatbots for routine tasks. However, human judgment and sensitivity remain crucial, as robo-advisers cannot truly care about customers.
Humans should control machines to ensure reliable and trustworthy financial services. Addressing challenges like bias, discrimination, and privacy requires regulation through an augmented intelligence collaborative approach.
Creative Art: An Autonomous Artist AICAN, an AI-driven art creation process, aims to generate innovative art rather than mimic established styles. Research reveals that AICAN's artworks are often mistaken for human-made art, indicating their potential for creativity. The collaborative approach between artists and AI recognizes AI as a valuable tool to enhance creativity. The evolving field of AI in art creation requires a broader understanding and appreciation of its creative contributions.
Fairness and Bias: The rapid advancements in AI applied to healthcare, medical diagnosis, and other domains raise concerns about fairness and bias. Biased AI systems can perpetuate inequalities in areas like healthcare, employment, criminal justice, and credit scoring. Bias, stemming from data collection, algorithm design, and human interpretation, causes unfair outcomes. Machine learning models may perpetuate data biases, leading to discriminatory results. Human-AI collaborations help to mitigate the bias in AI systems.
The Future of Human-AI Collaboration
Stakeholders must implement suitable AI management instruments to ensure effective human-AI collaboration and foster trust. Some of these may derive from legally formulated management approaches, while others should be innovatively designed as management by oversight. Institutions responsible for evaluating conformity assessments, registration systems, and post-market surveillance play a crucial role in ensuring credibility, especially at the national level, where enforcement and supervision of new rules lie.
The ultimate goal is to establish transparent conditions for human oversight of AI mechanisms and functionalities, facilitating fruitful collaboration between humans, machines, and AI teams. These adaptable conditions should serve as a constant foundation to benefit diverse human roles vis-à-vis AI systems, including doctors utilizing AI for personalized diagnosis and therapies, public institution officers developing automated decision-making processes, and city managers using AI for enhanced security solutions while safeguarding residents' rights.
Human-AI collaboration represents a promising future where humans and machines work hand in hand to achieve superior outcomes across various domains. By embracing collaborative intelligence, we can unlock the true potential of AI while retaining the essential human touch in decision-making and creativity. With careful consideration, transparent conditions, and innovative management approaches, we can forge a successful path forward into a world enriched by the collaborative synergy of humans and AI.
References and Further Readings
1. Boni, M. (2021). The ethical dimension of human–artificial intelligence collaboration. European View, 20(2), 182–190. https://doi.org/10.1177/17816858211059249
2. James Wilson, H., and Paul R. Daugherty. (2018). Collaborative Intelligence: Humans and AI are Joining Forces, Harvard Business Review.
- Inge Molenaar. (2022). Towards hybrid human-AI learning technologies, European Journal of Education, John Wiley & Sons Ltd. DOI: https://doi.org/10.1111/ejed.12527