Pulse Check: How to Make AI Feel Human, Safe & Transparent [Lightning Session]

0:00Hi, I'm Erin Newton, Content Director at Element 451. In this Pulse Checkup episode, I'm sharing highlights from my lightning session at the 2026 AI Engage Summit about human-centered AI design. We'll dig into how to design for trust, why it matters right now in higher ed, and what this looks like in practice. Let's jump in. Today, I'm going to be talking about designing for trust, how to make AI feel human, safe, and transparent. We have gone through a lot of different AI-powered tools and experiences these past couple days. Some of even what Devon was talking about with Discovery, as you're evaluating or maintaining these products, I want to give you some ideas of what you should be looking for to really make sure that they are going to engender a lot of trust with your users. It's really important for the credibility of your brand or institution, and this is for, you know, it's humans, all of us. This is very inclusive. These are things that staff, faculty, students, adult learners, people on the go on their phone.

1:06It's important to all of them. So let's start out where AI fails, actually. I've rounded up a few examples. This is Generative Search, so an AI overview. Somebody had a question about nutrition. How much is 20 grams of protein? And if you've ever met nutritionists or dieted and, you know, the ways that they kind of explain how much portions are, what they look like, one of them is a deck of cards. So the AI answered 20 grams of protein, looks like 3.5 ounces of extra firm tofu, 3.4 cups of cooked lentils, or a deck of cards. Nobody wants to eat a deck of cards, obviously, but you can see where the AI got that from because it was in the materials that it was generating its answer from. A little bit silly, not really a big consequence, but then you take a look at a couple of years ago, somebody was purchasing an airline ticket on Air Canada. And it was actually, there was a death in the family, it was a bereavement purchase.

2:06And so they wanted to make sure that they would be compensated for the ticket afterwards. So they asked the chatbot, what's your policy around this and the chatbot gave it an answer. The person actually screenshoted the answer. They went through about the ticket. All of that came time to process the refund to ask for it. And Air Canada said, nope, that's not actually our policy. Well, because they had the receipts, as the kids say, they were like, here, this is what your chatbot told me and went to court. And Air Canada was found liable. And we will get back to this story a little bit later, because it's the headline is little misleading. But all of this is a show, you know, these are examples of where AI, the trust fails. And how do we kind of engender some more trust? One of the things we can do is think about human-centered design. And you may have heard this term, and it may feel like something of a buzzword, but this is a longstanding discipline. This is a definition from IDO, which is more of a commercial consultancy who have brought, helped, you know,

3:12million-dollar companies approach their design through human-centered design, which really means having humans involved through the whole process. And it can be anything from a dishwasher to the remote control for your TV or software. And it's a lot of times involves multi-disciplinary teams, so you might have, you know, someone who's interested in color theory and knows how humans psychology reacts to certain colors. Or if you're making something, the materials that might feel, you know, a certain feeling or response. So that's human-centered design. And as you can imagine, that is now being applied to AI. So AI-powered experiences, you know, they can also be physical things like the, you know, the robots that we've seen in the past couple days that might be in our houses folding laundry. How do you take a human-centered approach to doing that? I want to leave you with human-centered design is a pretty big area, but there are three things that I think when it comes to AI that you can just really have in mind.

4:13AI feels untrustworthy when one of these things breaks. So transparency, tone, and accuracy. And so I've rounded up instead of the AI fails, some AI successes where we can look at where things are, these characteristics of human-centered design are really shining through. So first up is transparency. So, umo is an LLM that is encrypted, it's safe, it's secure. And this is, you can imagine the, the audience for this are people who are really concerned about their information not being used in training models, conversations, not being saved, not being stored. And it could be a variety of reasons. It could be, you know, financial information, healthcare. There's a lot of reasons why somebody would want a really locked down experience using an LLM. And so on the website, it makes it super clear right away. Hey, ask me anything. It's confidential right there in the headline. Super transparent. This is private, unlike other assistants.

5:13I don't record our conversations, safe guarded, treated with respect. Your conversations are never used for training. And another thing is that signals transparency here is the words they're using are really straightforward. They're not talking about databases. They're not using the term large language models. These are really straightforward words that, you know, you don't need to be a technologist to understand what they're saying. And why you would want to use this if security is really at the forefront of why, which kind of tool you want to use. Next up, we'll look at a different kind of AI powered experience. And this is Grammarly. And so this one is a really good example of some of the tenants of human centered design, particularly in kind of user experience digital experiences. Humans really want to know the why of something. So here, somebody wrote in Gmail, I think we should be able to solve this issue for you. And Grammarly says, do you want to sound more confident?

6:14Here is an edit that I suggest. It's saying the why to be more confident. It's very clear. You know, this is what they're suggesting. This is what would be edited. And then another really important tenant of the human centered design is to give the user clear control. So here you have accept or dismiss. It's totally up to the human. They could just say, nope, I don't want to do that. Or yes, I do. So they really are at the driver's seat. And that is really what can make something different than, you know, something where the AI is taking over. No, the human is still in control. They're making the decisions. The next characteristics of kind of how we can build for trust is with tone. And tone is more about the feelings of something that a person might have using. Whether it's an app or going on a website. And this is for lemonade, which is it helps people by renters by insurance for the rental properties. And you can see here, you know, there's an image of a person.

7:16There is very enthusiastic, positive language. And lemonade, you can tone is also really related to a brand and a brand experience. And you can tell they are really trying to signal this is a modern way to purchase insurance. It is not anything like what probably we have in our mental models of purchasing insurance, which is really kind of an off-putting, probably last thing anybody wants to do. Here, they've decided let's just be really friendly about it. And they take that further where, you know, instead of filling out a form, it's a conversation. It's, do you share your home with any of the following? And then you might, you know, they have these nice warm welcoming little drawings, roommates. And you can see the little help copy that comes up if you do select roommates. Your roommate's stuff isn't covered. They need to buy their own policy. So you can imagine in more of a traditional experience with an insurance company that would be like in the terms and conditions, tiny, fine print.

8:17This is right there transparent. And the tone is friendly. They're using your roommate's stuff. You know, these are very casual terminology. And what everybody would probably understand, like, oh, okay, I didn't know what stuff is. It's not like belongings or some really kind of off-putting kind of word choice. And then, you know, do you have dogs? Do you have pets and the cats? And it's just, it's a very warm kind of experience. And then we take a look at another one, which still very much Apple probably are the company that brought human centered design to the forefront. If you compare to kind of the Microsoft example and then just that really big cultural shift of like, oh, this is what an Apple product. This is like simplified experiences all about. So they definitely know human centered design. But you can tell the tone is really different here. This is I locked into my own profile. I see its mind. So that's a signal for me to trust it. And then it says, what's going on? So you can imagine this is something that maybe if you went to the doctor's office or the car repair place.

9:22A human would ask you just what's happening. And then there are some categories to choose from. And in this example, I selected repairs physical damage. And then you go and there are a couple other important kind of human centered design things that this is doing well. Again, there's two paths. You can select from a list from a topic. And you can see the words that they've used here for these topics are very like conversational and very what users would probably think of. Like my screen is cracked. You know, the back of the phone is damaged. These are not things that probably, you know, you they might have on the on the back end of how they would describe these issues. These are more kind of very human just everyday terminology that people would use. And then they have done at the bottom or just tell us in your own words. So again, giving the user the ability to choose their path and how what works best for them behind the scenes. So AI powered. But to the person on the other end, you know, there's two paths that are pretty clear. They're in their language. So it's just a very supportive kind of experience.

10:29And then the last bucket here going back to the air Canada stories accuracy. And so I mentioned that the headline here was a little misleading. The chatbot didn't exactly make up the policy. There was a policy for the particular bereavement airline ticket policy that they had it on their website. That was the correct one. And then the chatbot was either confused didn't know which knowledge base to find the information from could have been who's nation or just it chose the wrong thing. So really where this is going where us as communicators marketers working in higher ed that the skill set is going to change a little bit from, you know, writing and generating content to really governing content. So accuracy depends on having a clear defined knowledge base. So Devon, how we showed the discovery tool, you know, there's a knowledge base behind that whether it's your website or it is files. And that's really where the time and energy needs to be spent a lot to make sure that all of that is accurate so that your whatever AI is, you know, giving to the user on the front end.

11:40And it reduces the chances of hallucinating if it knows the right things to say and it's not enforced in these positions to just make things up because it's confused. You also want to really think about your high risk content. There's so much content. So really I prioritizing one of the things like deadlines, anything financial, anything where like you're going to kind of mess up somebody's big plans for their life. If that information is not accurate and it reflects poorly on your brand, it's just nobody wants that avoiding conflicting sources of truth. So if you've worked in higher ed for a while, websites, notorious place where there's a lot of content owners. It's really hard to wrangle all of that, but that again is going to be a skill set where, you know, people are going to thrive who are able to manage content and govern it. So that the AIs are working and presenting the right information. And then the other thing is just know when an AIs not the right person to answer to deliver that trustful experience and guidance.

12:43There are certain things that an AIs just not meant for that we really want to bring the human in. So that I hope you keep all of these three things in your mind when you're either building or buying an AI powered experience for your staff, faculty, students, yourself, transparency, tone, and accuracy. Thanks for listening to this Pulse Check episode. If you're continuing the conversation around human centered AI, I'd love for you to take one idea from this episode and put it into action on your campus or workplace. Thanks again for tuning in. Thanks for listening to The Higher Ed Pulse. The show is a proud member of the Enrollify Network home to higher ed's largest collection of podcasts. Enrollify is where higher ed goes to grow. So be sure to check out our other podcast webinars, video clips, blogs, and more.

13:43Visit Enrollify.org and subscribe for all the latest details and content drops directly to your inbox. Enrollify is brought to you by element 451, the digital workforce platform for higher ed. Learn more at element 451.com. Thanks for listening to The Higher Ed Pulse.