Editor’s note: WRAL TechWire contributor Dr. Sarah Glova is a globally recognized speaker, successful entrepreneur, university instructor, and business consultant. A seasoned educator and entrepreneur, Sarah is CEO of the award-winning digital media firm, Reify Media. With a Ph.D. in Instructional Technology and a Master of Science in Technical Communication, she is dedicated to cultivating forward-thinking work environments.
Her new column asks industry leaders to share their insights, experiments, and concerns about AI in their respective fields.
+++
RALEIGH — Usability expert Dr. Kenya Oduor has a helpful metaphor for how we assess artificial intelligence (AI).
To her, it’s like dating.
“It’s a reciprocal, interpersonal type of relationship,” said Oduor. “And, just like when you get to know someone, that relationship progresses over time. The same thing happens when we interact with things like AI.”
Sarah Glova
Oduor is an adjunct professor of computer science at NC State and founder of the local human-centered design and research agency Lean Geeks. She spoke with me about “Natural AI,” or AI that can “communicate and engage with us, as humans, in a way that feels intuitive, feels human-like.”
And she says the experience of “interacting with those technologies” — specifically, deciding to “trust and rely on” those technologies — is very similar to dating.
With so much AI growth on the horizon—we’ll be meeting some new bots
AI is expected to have a “rapid” annual growth rate of 37.3% between 2023 and 2030, Grand View Research reported in February.
And the demand for these tools to feature “Natural AI” is growing, says Nazneen Ahmad, in a blog post written for LambdaTest, the cloud-based testing platform.
“The demand for chatbot-driven interactions is on the rise, and advancements in AI technology are making chatbot responses more natural and human-like,” writes Ahmad.
As people encounter new AI tools, Oduor says they will “evaluate” how effective these tools are pretty quickly — the same way people “evaluate” a first date.
“Just like you would with any other system, you’re evaluating its effectiveness,” said Oduor. “How many times do I have to ask the same question, or in how many different ways, in order to get the expected assistance or response? So if I have to keep repeating myself — versus one time and I get what I need to help me move through my task — that’s when you know if it’s helpful or not.”
TechWire columnist Sarah Glova cracks ‘2023 40 Under 40 Awards’
What if the date isn’t going well?
I asked Oduor — what if the Natural AI doesn’t live up to expectations? How much tolerance do people have for a “bad first impression,” so to speak?
She told me — it depends on the errors.
“If you think about the dating scenario, and the difference between — you showed up late, or you left me at the restaurant because you were a complete no-show,” Oduor explained. “I might have some tolerance around, ‘You showed up late.’ If you’re a no-show, that’s a complete deal breaker.”
She said that, similarly, people might have a tolerance for some errors—but there’s a threshold. Some errors are a “deal breaker.”
“And so when you look at the extent, or the impact of the undesired response from the person you’re dating, that’s where you assess whether or not I’ll continue to give them more time, and another chance, versus ‘I’m done, and I’m going to move on,’” said Oduor. “So I think in that evaluation, you’ve got to consider — what is the extent of the error or the inability to help me move through my task or accomplish my goal.”
She added that users aren’t looking for perfect AI—they’re looking for transparency.
“With AI in general, the ability to demonstrate transparency, but also your level of confidence,” said Oduor. “So you can ask me about something factual from history, and I can tell you what I think, or what I’ve heard, but if I’m not an expert, there might be some holes or gaps in what I’m sharing with you — and I’ve got to be able to say, ‘You know, I’m not an expert on that, but this is what I do know.’”
She said that transparency is more important than the AI getting the right answer all the time.
“Even if it’s not 100% reliable, can they still develop trust? Yes, because you’re telling me, ‘With this level of confidence, I’m giving you this response,’” said Oduor.
What can companies do to encourage trust and transparency? Play the dating game.
As always, I asked ChatGPT for an opinion — with the prompt, “If we can compare meeting a new AI tool to dating (with the idea that when a user meets a new AI tool, the user evaluates it the way a person might evaluate someone on a first date), then what can companies do to help ensure their AI makes a good impression?”
The response focused on a general description of AI that’s reliable, knowledgeable, personalizable, and transparent:
“To ensure their AI makes a good impression on users, companies can take several steps. First, they should focus on creating a user-friendly and intuitive interface, allowing users to easily interact with the AI tool. Clear and concise instructions or prompts can help users understand how to engage with the AI effectively. Second, the AI should be well-trained and knowledgeable in its specific domain, providing accurate and relevant responses. Thorough testing and continuous improvement of the AI’s capabilities can help ensure its reliability. Third, companies should prioritize personalization and adaptability, tailoring the AI’s behavior and responses to individual user preferences. By understanding and addressing user needs, the AI can establish a sense of connection and rapport. Finally, maintaining transparency about the AI’s limitations and capabilities can foster trust and manage user expectations, avoiding potential disappointments.”
– ChatGPT inquiry, June 15, 2023
But Oduor gets the last word.
I asked Oduor what companies can do to prepare their AI tools for the “dating scene.” She recommended that teams focus on the dating metaphor.
“Your users or consumers are not thinking about, ‘Oh, this is like dating,’” said Oduor. “But if you want to put yourself in their shoes, you want to understand their context, then you also want to understand that they’re looking at this relationship like dating.”
She also recommends the book Rebooting AI: Building Artificial Intelligence We Can Trust, which she has assigned as required reading in her courses.
For more about Oduor, visit her company’s website at leangeeks.net.
