Expert Panel | Forbes.com »
Artificial intelligence has seen a meteoric rise in recent years. From chatbots and marketing initiatives to automating workloads and improving customer service, businesses are increasingly turning to AI to streamline operations, while consumers use AI-driven tech like smart speakers to make their lives easier and more enjoyable. As the technology behind artificial intelligence evolves, developers must continue to consider one of the primary goals of AI development: to make it more “human” than “artificial.”
More human-like AI is not only more likely to be trusted and adopted by the general public but also provides a smoother, more successful user experience. If you’re working to build an AI application that’s more “human-centric,” keep these 15 important factors shared by the members of Forbes Technology Council in mind.
1. Remember that humans will ultimately manage the application.
In the development of AI-driven products that engage humans by way of interaction or emotion, always remember that humans will manage the application. The learning and optimization by human intelligence should be leveraged over fully autonomous design and determination by artificial intelligence, as humans know humans better. – Dana Ghavami, Spotible
2. Give it a human name.
I have experience building AI chatbots, and we eventually realized we had to give the AI a name to make it more “human.” For example, we called our event search chatbot “Eve,” which is short for “event.” People remember names, and they will have an emotional connection to the AI if it is not called by the function it performs but is rather “personalized.” – Bobir Akilkhanov, Missed.com
3. Treat AI like a team member.
Make the AI part of your team. Just like a human team member, you should train it, performance-manage it, guide its ethics, give it a voice and improve its learning. That’s how you get value, acceptance and human-centered AI evolution. – Antonluigi Gozzi, LiveHire Ltd.
4. Make the technology effortless.
Technology should fade into the background. The interface needs to be effortless for a human to naturally interact with the AI. Further, ensure robustness, especially if it’s released to the public at large—end-users will attempt to troll or trick the AI and/or lead it to edge cases where it cannot recover. Account for the usage of foul language, colloquiums, etc. – Sreenivasan Iyer, Shasta Ventures
5. Ensure it’s simple to understand.
Regardless of the level of automation, there’s always someone on the other end interacting with the AI application, so we need to make it simple to understand. If it’s being used to gain business insights into patterns, creating easy-to-understand reporting tools goes a long way. Professional communications input is crucial if you’re going to hammer down phrasing, interactivity and visualization. – Arjan Wijnveen, CVEDIA
6. Provide context.
Explainability in AI plays a critical role in building more human-centric systems. AI should not be a black box that presents outputs without the context that humans need to use those outputs. Findings should be provided along with rationale so that humans understand the “why” rather than just the “what.” Explainability can prevent humans from thinking of AI as an oracle shrouded in mystery. – Dhananjay Sampath, Armorblox
7. Include the ability to escalate to a person.
AI is software that learns to pick up patterns. It’s designed to improve as time goes on, just as humans do. But AI isn’t designed to stand alone. When building an AI application, it’s critical to consider how the technology gracefully escalates to a person when it hits a roadblock. This human-in-the-loop concept means that AI can do what it does best while still providing a stellar experience. – David Karandish, Capacity
8. Make the interface intuitive, personal and friendly.
Routine decisions are mainly intuitive, gut-feeling decisions that are prone to human biases. Good, bias-free AI decisions often seem counterintuitive, leading to mistrust and resistance. This calls for a user interface that is intuitive, personal and friendly and which explains the AI decisions, allowing for human overwrite. Monitor its performance to gently teach and convince the user to trust the system. – Michael Feindt, Blue Yonder
9. Collect human behavior data before developing a solution.
One important factor to remember is that humans can be both very predictable and unpredictable. If the application is responsible for monitoring human activity and innately responding to something it observes, the team must understand and study the human behaviors they are dealing with. Once enough “human behavior data” is collected, they can begin to develop a suitable human-centric solution. – Sandeep Pandya, Everguard.ai
10. Ensure your team has adequate skills to scale AI.
Scaling AI requires the right skills. Companies often focus only on scarce tech talent such as data scientists. That’s crucial, but for human-centered AI, you need up-skilling and cross-skilling so that all team members are speaking the same language and working together. Business users need to know enough about how AI works, and data scientists need to know enough about your business to design smart solutions. – Matthew Lieberman, PwC
11. Include the ability to recognize and respond to different situations.
Artificial intelligence covers a broad range of applications, from determining the optimal routes across a network to recognizing investment trends to appropriately answering a user request. In the context of a live person interacting with the application, the challenge is delivering an application that’s good at recognizing when the user wants the AI to “do what I meant, not what I said.” – Saryu Nayyar, Gurucul
12. Test and improve the conversation flow.
The key to making your AI more humanlike is to test and improve the conversation flow. Chatbots are great for helping customers with simple issues, but the delivery can significantly impact UX. Spend some time going through your AI system and fine-tuning the way it responds to users when they have specific questions to create a more traditional customer service experience. – Thomas Griffin, OptinMonster
13. Give it a personality.
People tend to remember how you made them feel. Unfortunately, most AI feels robotic, which means you don’t feel much when interacting with it. Want to know why Siri became a cultural phenomenon that’s now embedded into every single iOS device? Because it is witty and snarky. When programming AI that interacts with people, give it personality, charisma and empathy. – Marc Fischer, Dogtown Media LLC
14. Add diverse data.
Consciously make an effort to add some diversity of thought and data. Bias has been creeping up lately as a major problem in AI. It’s like a baby—it learns what it sees and hears. If you’re feeding the system base data, make sure to fact-check the sources, and try to have as many diverse sources as you can. If it learns from human interaction, you should include people from diverse backgrounds. – Vikram Joshi, pulsd
15. Support human feedback and auditing.
Data-driven decision-making can help protect from certain hidden biases in human intuition, but AI has its own limitations that must be managed. Human-centric AI applications must proactively protect against the model’s overfitting against training data. They should support human feedback and feasible auditing to decide when underlying observations necessitate model adjustments. – Andrew Sellers, QOMPLX, Inc.