How to make AI more "human"? MIT research: This is crucial

How to make AI more "human"? MIT research: This is crucial

Written by: Green Apple

Editor: HS

Layout: Li Xuewei

As more and more technologies are developing, machines are interacting with us in a more natural way. Among them, Voice User Interfaces (VUIs) are a major way to achieve social interaction.

Now that voice assistants have become standard equipment for technology giants, the next step is to bring each family member closer to this AI assistant.

A new study from MIT Media Lab shows that users are more likely to trust new VUIs if the digital assistant exhibits some human-like social behaviors. More specifically, the more social behaviors the VUIs exhibit, the more they are in line with human wisdom and habits, the more likely people are to trust it and interact with it.

The research results were published in the journal Frontiers in Robotics and AI under the title “Speed ​​Dating with Voice User Interfaces: Understanding How Families Interact and Perceive Voice User Interfaces in a Group Setting”.

In the paper, the authors propose a speed dating scenario that simulates family members trying different VUIs assistants, providing a basis for users to develop a mental model of virtual terminals, which can be used by relevant practitioners to understand users' initial perception of these devices.

Human family members generally tend to think that a device is more capable and emotionally engaging if it can detect and display social cues, such as looking toward someone who is speaking.

Additionally, the study revealed that branding effects—particularly whether the manufacturer’s name is associated with the device—has a significant impact on how family members perceive and interact with different VUIs.

Of particular note, when the device had higher levels of social expression, such as the ability to provide verbal and nonverbal social cues through movements or expressions, family members interacted with each other more frequently while using the device.

These findings could help product managers and hardware designers create VUIs that are more engaging and more commonly used by family members, while also improving the transparency of these devices.

"The home is made up of family members, so we were interested in looking at this from an intergenerational perspective, including children and grandparents," said Anastasia Ostrowski, an author of the paper. "It was very interesting for us to understand how people perceive these devices and how family members interact with them."

Investigate the interactive process

The work stems from a study that explored how people use VUIs in their homes.

In a series of three studies on social embodiment, the research team iterated on a speed dating format of elicitation studies to provide insights into user interactions and user perceptions of commercially available products and the robustness of these results, and to demonstrate how research can be iterated and refined to support the replicability of concepts.

Before the study began, users were familiarized with the three devices and then took them home for 1 month.

So why do people prefer social robots?

To find out, they designed three experiments, with one group from each family interacting with different VUIs. Thirty-four families with a total of 92 members participated in the study, ranging in age from 4 to 69. Each study included three activities to examine participants’ interactions and perceptions with three VUIs, including an agent exploration activity, a perceived personality activity, and a user experience ranking activity.

The experiments were designed to simulate a family's first exposure to VUIs. During the experiment, all families were required to interact with the three devices, complete a total of 24 actions (such as "asking about the weather" or "trying to understand the agent's advice"), and record the video. Then, they answered some questions about their views on the devices and were classified according to the personality characteristics of the VUIs.

In the first experiment, we explored how users interacted with and perceived three commercial devices without modifying their commercial default settings.

Unexpected results

In a second experiment, investigating the same three VUIs, the researchers set out to understand how branding influenced participants’ opinions. In addition, they provided participants with information about each manufacturer.

Changing a device’s wake word could have ethical implications. A personified name would make a device seem more social, she said, but it could also mislead users by obscuring the connection between the device and the company that made it, which now also has access to the user’s data.

In a third experiment, the team looked at how interpersonal motion affected interactions, suggesting that repetitive motion without social embodiment might not be an effective way to increase user engagement.

Each study revealed strikingly similar patterns of results, demonstrating the impact of concepts such as embodiment and social presence on user interactions and perceptions.

Cultivate deeper relationships

A deeper analysis of the third study showed that when the devices they used had more social capabilities, users interacted more, such as glancing at each other, laughing together, or chatting with each other.

“For families, we’ve always wondered how these systems can facilitate interactions between users,” Ostrowski said. “That’s always been a big question: How are these devices going to shape relationships? We want to design systems that can help people connect more closely.”

Using their insights, the researchers propose several considerations for the design of VUIs, including the importance of cultivating a warm, outgoing, and thoughtful personality; understanding how wake words influence user adoption; and conveying nonverbal social cues through movement.

With these results, the researchers want to continue exploring how VUIs with different levels of functionality can be used in the home. For example, they might conduct studies with three different social robots. They also hope to replicate these studies in real-world settings and explore which design features work best for specific interactions.

Perhaps one day, smart assistants will really become an indispensable member of each of our families.

Academic headlines

<<:  Why does winter make people want to fall in love?

>>:  Winter Olympics mascot: I was almost able to recognize the prototype!

Recommend

Google Lens feature added to the Google APP on iOS

[[252266]] In March this year, Google Lens was la...

How to recall lost users? Analysis from 3 aspects!

Some people say that whether or not one can succe...

6 ways to monetize community operations

I always encourage you to build communities becau...

iOS 16.5 push update, the official version will be released soon

iOS 16.5 Beta 4 Update Yesterday morning, Apple p...

What is the difference between Douyin Box and Douyin?

Each friend chooses a different way to play on th...

The underlying logic of brand growth and SOP implementation steps

In terms of underlying logic and final results, t...

Google Glass can be used to steal passwords with 90% accuracy

The latest research results from the University o...

The rise of HTML5: rejecting aloofness and niche

Shikhir Singh is currently working at Sencha as a...