When artificial intelligence was in its infancy, it was anticipated that someday it would function without human prescription. Several decades later, while the technology has evolved, this is not yet the case.
This is AI’s big secret – it still doesn’t work without humans.
Whether on the input side, where the AI is being programmed, or the output side, where it may require human interaction to finalize or execute outcomes, AI is typically dependent on us to work effectively.
The story of AI is inextricably linked with humans. For organizations making use of the technology, the most important consideration is whether AI is positively impacting the people who interact with it.
A human-centric approach to AI solutions
When considering a new AI implementation or evaluating an existing one, there are two major aspects to consider. There’s the technical track (specific architecture, rollout, costs, and functionality) and the human track. A common mistake for organizations is to rush into developing the technical without considering the human.
An effective way to remedy this is to build project teams with both field researchers and technicians. The field researchers conduct interviews to understand the personas and scenarios of everyone affected by the new technology. The work of both the technical team and field researchers should converge at the final stage of the planning project for the overall design.
With AI focused entirely on delivering specific human outcomes, companies can avoid unnecessary spend and significantly improve project ROI.
Does human involvement invite human flaws?
A common argument against AI in general is that it perpetuates systemic biases that exist in the historical data upon which it is trained. The evolutionary need to make impulsive judgments based on prior experiences initially served to increase the likelihood of an individual’s survival, or at least a positive outcome in a given situation. But when prior decisions and outcomes are used to train functional AI without context, it risks being wrong.
Subconscious biases held by developers of AI could unwittingly introduce ageism, ableism or racism to the technology so it ends up mirroring the prejudices of society. But the issue around biased AI can be remedied with effective governance. Rather than attempting to gatekeep inputs or outputs, this involves an entirely different mindset for design.
Governance should be an embedded capability across the technology stack – a platform that sits horizontally across all stages of the architecture, not just the beginning or end.
Take the example of a bank attempting to evaluate whether an individual is credit-worthy. With AI, there’s no need to take a rules-based approach such as scoring by postal codes, which will almost certainly lead to unfair outcomes.
Instead, AI can make nuanced judgments based on all of the factors available. It’s impossible for a human brain to evaluate all these dimensions at once and come to an optimized recommendation.
Will AI negatively impact humans by taking jobs?
In the eighties and nineties, it was commonly envisioned that AI would eventually replace manual roles.
But this hasn’t happened to the extent it was predicted. You see as many humans as ever on a modern car assembly line. Somewhat surprisingly, professional roles have been affected by AI to a comparable, if not greater extent. Professions such as medicine, law, and even data science itself require skills that are greatly accelerated, if not somewhat replaced by good AI.
Of course, trained data experts are still required, but outcomes can be optimized and scaled through AI. Take Uber, for example. As often reported, the fundable concept was that autonomous vehicles would be ferrying humans from place to place by 2022, while the capitalists behind those vehicles would be raking in the dollars and masterminding the operations. Instead, AI acts as the central brain handling route identification and optimization for human drivers. So humans are responding to an AI boss, rather than the other way around.
More generally for AI automation, it was believed that humans would be in the driver’s seat for more creative problem solving, while AI handled most of the leg work. So perhaps the main concern around the relationship between AI and humans in the workplace is less about them being replaced, and more about ensuring they remain creatively fulfilled in their role. We need to balance short-term gains in efficiency with long-term sustainability of labor systems that work for everyone.
In short, the workplace relationship between humans and AI comes down to purpose. Not only must the AI solution be fit for purpose, but it must also deliver outcomes that align directly with the purpose of people within the organization – augmenting their capabilities and efficiencies to deliver positive results for all. AI that can work alongside people in this way is more important than any standalone functionality.