The science is clear.

Diversity programs often fail because they don’t create the necessary pathways for inclusion. Studies by Frank Dobbin and Alexandra Kalev confirm that behavioral change begins when we address our biases openly, and effective social contact reduces our biases. In practice, individuals that work side by side begin to see each other as equals. Familiarity fosters acceptance because we find the things that unite rather than divide us.

In a recent study on growing at work, scholars Scott Sonenshein, Jane Dutton and Adam Grant found that employees’ personal perceptions of their own growth were deeply embedded in social context: participating in group achievements, learning a common language, learning from colleagues and helping colleagues learn. They attributed these achievements to communication, mutual support and the creation of a “safe space to take the time to acquire new perspectives.”

In a compelling study at MIT, Alex Pentland used bodily-worn sociometers to measure social interactions in a former Bank of America call center. These digital devices measured the length and depth of interactions among colleagues. Colleagues who interacted more had faster calls, less stress and the same approval ratings as their colleagues who stayed glued to their phones. When the company synchronized employee breaks across their call centers, they saw a 20% drop in average call handling time—even for the lower-performing teams—and a $15 million improvement in productivity over one year. It may seem counterintuitive, but it’s powerful: Taking time to connect with colleagues fosters increased productivity.

Interpersonal relationships have long been identified as a key driver of organizational learning in the R&D space. In a landmark study, management scholars Wesley Cohen and Daniel Levinthal coined the term “absorptive capacity,” an organization’s ability to recognize and apply new knowledge and principles. Organizations with the highest absorptive capacity explored and converted new ideas by supporting individuals’ unique knowledge and experience and creating space to share diverse perspectives. Social connections unlock these ideas and put them to work.

Science traditionally has accepted a general intelligence factor as a measure that gauges and predicts individual performance. But a recent study, led by Anita Woolley at Carnegie Mellon with colleagues at Union College and MIT, found evidence of collective intelligence, a new measure to reliably predict team performance. The “smartest” teams are not those with the highest average individual intelligence, or even those with the highest maximum individual intelligence. The smartest teams have higher social sensitivity and balanced conversational turn taking–they treat each other with empathy and no one dominates the conversation. And they are diverse: these teams have a higher proportion of women.