Creativity is powerful. Regarded as one of our most sophisticated cognitive skills, this ability drives human progress by allowing us to perform nonroutine tasks, take advantage of novel opportunities, and invent new solutions to problems facing the world. Take plant-based meat substitutes: Using new flavor science techniques, researchers have been able to create products that taste like meat so as to take advantage of emerging health and sustainability opportunities. Creativity isn’t just reserved for the culinary arts, of course; it’s also the hallmark of engineering and technology that benefits wide swaths of society. More broadly, it’s the basis for innovation and continuous reinvention—enhancing creativity can speed up and improve product and service development across all industries.
According to psychologists, creativity is the generation of an idea or artifact judged to be both novel and valuable by a suitably knowledgeable social group (for example, recipes that are both different from anything ever seen in cookbooks and amazingly flavorful). I agree that creativity should be judged on these two socially constructed dimensions, though I caution that some transformative creativity may be beyond the ability of social groups to assess. I’ve even mathematized this two-dimensional definition to develop a mathematical theory of creativity that establishes a fundamental limit to distinguish what’s possible from what’s impossible in a given creative domain. Rather than comparing creativity to previous ideas or artifacts, such limit theorems define fundamental benchmarks and can push technologists to be more creative so as to approach the absolute limits. Those technologists might then go on to adopt creative processes they might not have otherwise.
Although we still lack a full understanding of the informational, computational, psychological, or neurological underpinnings of creativity, we’ve come a long way from when it was considered an unintelligible and mystical power. Scientific investigations, for example, have elucidated that creativity is a specific cognitive process that can be influenced by a wide range of factors including personality, motivation, and environment. Simultaneously, advances in artificial intelligence have enabled engineers to build systems that achieve creative performance at human —or superhuman—levels, not just in artistic domains but in scientific and engineering ones as well. Examples from my own research include algorithms to generate surprising and high-quality recipes, unit processes in engineering, and sustainable building materials that are stronger than existing formulations despite requiring half the carbon emissions to make. Other researchers have developed systems to design quantum information processing circuits, proteins to serve as antibodies for novel diseases, novel metal alloys, and the list goes on!
Popular culture tends to lionize the lone genius but group creativity often trumps individual creativity.
Although popular culture tends to lionize the lone genius, group creativity often trumps individual creativity, especially in the face of unexpected or extreme societal challenges. Perhaps the most well-dramatized version of this is the creativity shown by a group of engineers in saving the Apollo 13 mission. As such, it’s not surprising that technical and engineering work is largely carried out by teams rather than heroic solo inventors. Teamwork, however, comes with its own dynamics. Effective collaboration requires both cooperation, i.e., having aligned goals, and coordination mechanisms to enable effective alignment and adjustment to teammates’ actions. Both are enhanced in social creativity if teammates have theory of mind, or the ability to attribute mental states such as beliefs, desires, and emotions to oneself and to others. Appropriate social workflows and organizational architectures also help.
How effective are existing remote work tools at facilitating creative collaboration from a distance? The COVID-19 pandemic has placed an increased emphasis on this question as technical and engineering teams have largely shifted to remote work, an arrangement that seems to be here to stay, at least in part.
How effective are existing remote work tools at facilitating creative collaboration from a distance?
As a result, videoconferencing, social networking, digital curation, and education tools are all seeing significant increases in usage as teams interact virtually. Unfortunately, these tools seem to be ineffective at fully recreating the wide range of social heuristics and institutions that support creativity in shared physical space. These shortcomings have a deleterious effect on both productivity and work satisfaction. A recent study conducted by Steelcase, an office productivity company, suggests that since the pandemic, time spent on routine, individual task work is up, while time spent on collaboration and creative tasks is down. When Wired asked Google and Alphabet CEO Sundar Pichai about his company’s adaptation to remote work conditions in May, he responded with a question that underscored his concern about remote creativity limiting productivity: “How productive will we be when different teams who don’t normally work together have to come together for brainstorming, the creative process?”
I believe AI-based systems can bridge this disconnect. For example, information lattice learning techniques for knowledge discovery may be capable of facilitating co-creativity among technical and engineering teams far better than any currently available remote work tools. The basic idea is that AI—whether it takes a physical form or not—can support human cooperation, coordination, and, ultimately, creativity by taking on various roles, such as that of “nanny,” “pen pal,” “coach,” and “colleague,” as part of a human-AI creative network. In this context, one of the AI agent’s critical roles is to sustainably synthesize individual contributions into high-quality creative outputs. Let’s say team members are remotely building new software. Each member of the team creates individual pieces like data access components, business entities, and service interfaces, while the AI system creates the “glue” that makes everything fit together harmoniously. In group music composition, this AI glue could bring together melodies contributed by two people, a rhythm contributed by another, and a musical style contributed by a fourth into a single unified piece—fitting components together and adding “wrapper code” of new notes that are needed to make it sing.
Because individual contributions to creative tasks are fundamentally complex and interdependent, combining contributions requires a higher degree of intelligence than a simple ensemble method for aggregation. How to aggregate disparate contributions in complicated domains such as engineering design has, however, been an open question in the fields of human-computer interaction and collective intelligence.
The AI-based aggregation glue I described is more than a single mathematical function that sums up or integrates individual contributions. Instead, it’s an ecosystem—a kind of metafunction—which creates, updates, selects, and manipulates algorithms desirable for co-creation in remote and distributed settings. The aggregation function combines multiple individuals’ contributions—their ideas, style, and interests—with the capabilities of AI agents, including algorithms, learning models, data infrastructures, development tools, and environments. Moreover, the aggregation function intelligently fuses contributions from these agents to enable strong creative reactions that yield high-quality results.
Such aggregation also does more than collaboration tools, which only bring people together and provide channels to talk (e.g., Skype for videoconferencing), sing (e.g., Smule for co-performance), or teach (e.g., Seesaw for online education). Instead, the AI system behaves as a unifying force—well beyond a communication channel—that plays a creative role in the activity itself by entering the creative conversation to facilitate and inspire people to co-work, co-contribute, and co-create. It stages activities for people so the final creative result is as high-quality as possible.
Such AI-based co-creativity systems would have far-reaching implications for remote work and beyond. By providing a cohesive glue, these platforms could enable cooperation among remote teammates at levels that potentially exceed those possible within in-person work environments. AI systems could unleash the power of collaboration in creative domains where it hasn’t traditionally taken hold, such as the visual arts. It could also empower people with highly specialized skills, such as a musician with exceptional taste in jazz reharmonization, to create a complete, cohesive work by collaborating with an AI system. Since creativity requires both generation and discrimination, there is scope for humans and AIs to potentially contribute to each. Creative fields manifest these roles in distinctive patterns, such as critics, galleries, and artists who produce for them. Reconsidering the scope of roles on technical and engineering teams could also be part of the design space for an AI-augmented system. Perhaps there is a new role for software critics waiting to be defined, distinct from quality assurance?
Beyond short-term productivity gains in creativity through improved coordination, AI-based group creativity systems could enhance employee satisfaction, which has a long-term impact on creativity. In psychology literature, self-determination theory states that people across cultures have three innate psychological needs: mastery, connectedness, and autonomy. Human autonomy is defined as having flexibility and control over processes and outcomes. Satisfying these needs is said to lead to psychological health and well-being, which in turn leads to greater creativity, effective problem-solving, motivation, performance, and persistence. When these three needs are satisfied, people experience jobs as self-determined, and their behavior becomes intrinsically motivated by personal goals and values rather than extrinsically motivated by external reinforcements or demands.
Following ethical principles of technology design may lead, in the long run, to greater group creativity.
Importantly, self-determination theory does not consider the three psychological needs to be equal in the experience of self-determination: Mastery and connectedness are conceptualized as part of the contextual background that gives rise to self-determination, whereas autonomy is viewed as the key ingredient. While existing remote tools often satisfy one need, group creativity activities satisfy all three at once, yielding the highest levels of health and well-being. For these benefits to accrue, however, AI-based creativity systems should be able to combine disparate contributions while retaining the distinct creative contributions from each person so as to maintain a sense of autonomy. In ensemble composition, melodic contributions by people should not be overwhelmed by the harmonization of AI so they cannot be heard. Unfortunately, cognitive support tools have a tendency to change content for creative activities and reduce the feeling of human agency, which people indicate is important to feeling ownership of creative artifacts.
Respect for human autonomy is one of four key principles in standard biomedical ethics frameworks and has been brought to technology design ethics and incorporated into several AI ethics frameworks. As the coauthors of the paper “Designing AI for Social Good” wrote, “It is essential that software intervenes in users’ [lives] only in ways that respect their autonomy. Again, this is not a problem that arises only with AI-driven interventions, but the use of AI introduces new considerations.” In other words, following ethical principles of technology design may lead, in the long run, to greater group creativity.
Many observers, myself included, believe that even after the COVID-19 pandemic abates, many of our work lives are going to be more physically constrained and virtual than before. To make this shift while maintaining productivity and well-being, we must ensure that remote work practices positively impact group creativity. At the moment, no available tools or platforms support co-creation of the type needed to combat the isolating impacts of social distancing. Yet networked artificial intelligence has the potential to combine the creative contributions of each individual on a technical or engineering team in a way that maximizes quality while still respecting each human contributor’s autonomy. The end result? An AI aggregation agent that, by bringing people and their contributions together, is capable of facilitating creative breakthroughs and the creation of new technologies that have never been imagined.