How to design and develop AI systems so cool that anyone would love them

Introduction

Have you ever wondered why everyone likes ChatGPT or why it have gained over 2 million of users in just a few days, which it’s totally a record break for any platform in the world.

Well, I have worked on a bit many artificial intelligence projects, especially on computer vision and natural language processing, but in the end they didn’t make it up as I expected, especially when I comes to the number of people who used them frequently.

I had come to realize that I wasn’t the only victim of such experience, many startups and companies have built many artificial intelligence projects, also especially on computer vision and natural language processing, but in the end they didn’t make it up as they expected, also especially when I comes to the number of people who used them frequently.

I then later came to realize that having an artificial intelligence project is more than having fancy GPUs or even powerful servers or computers, also on the other hand it’s not even about having powerful and highly optimized artificial intelligence models.

Since the systems were designed by humans in order to be used by humans, we forgot that and that’s why many of our systems didn’t work out.

When I applied these principles, the results were astonishing and left me to realize why I didn’t knew this from the first place.

And also I even came to realize that companies like Google, Microsoft and OpenAI use the same principle of which we take for granted that’s why even you can testify that it feels good to use their Artificial Intelligence systems.

The following are some of the guidelines to be followed when designing and developing Artificial Intelligence systems that are human centric.

The guidelines are divided according to the situations at which humans interact with Artificial Intelligence systems, this includes:-

1. Initially when a person interacts with the system for the first time.
2. During the interaction with the system after a while.
3. When something goes wrong during the interaction.
4. Overtime when the user gets highly familiar with the system.

Guidelines for Human-AI Interaction

Now let’s dive in to get a more understanding of the concepts:-

1. Initially when a person interacts with the system for the first time

a) Make clear what the system can do by making users understand what the system can do.

b) Make clear how well the system can do what it can do by making the user understand how often the system succeed or make mistakes.

2. During the interaction with the system after a while

a) Time all services based on their context by identifying the appropriate time when to act or interrupt based on the user’s current task and environment.

b) Show contextually relevant information by displaying the information relevant to the user’s current task and environment.

c) Match relevant social norms by ensuring that experienced is delivered in a way the users would expect, given their social and cultural context.

d) Mitigate social biases by ensuring that the system’s language and behaviors doesn’t reinforce undesirable and unfair stereotypes or biases.

3. When something goes wrong during the interaction

a) Support efficient invocation by making it easy for a user to invoke or request the system’s services when needed.

b) Support efficient dismissal by making it easy for the user to dismiss or to ignore undesired system services.

c) Support efficient correction by making it easy for the user to edit, refine or recover when the system is wrong.

d) Scope services when in doubt by sometimes engaging in disambiguation or graceful degrading the systems services when uncertain about a user’s goals.

e) Make it clear what the system did that by enabling the user to access the explanation of why the system behaved as it did.

4. Overtime when the user gets highly familiar with the system

a) Remember recent interactions with the user by maintaining short term memory and allowing the user to make efficient references to that memory.

b) Learn from the user’s behavior by personalizing the user’s experience by learning from their actions overtime.

c) Update and adapt cautiously by limiting disruptive changes when updating and adapting the system’s behaviors.

d) Encourage granular feedback by enabling the user to provide feedback indicating their preferences during regular interaction with the system.

e) Convey the consequences of user actions by immediately updating or conveying how user actions will impact future behaviors of the system.

f) Provide global controls by allowing the user to globally customize what the system monitors and how it behaves.

g) Notify users about changes by informing the user when the system adds or updates its capabilities.

I hope that you can use these Guidelines for Human-AI Interaction throughout your design process as you evaluate existing ideas, brainstorm new ones, and collaborate with the multiple disciplines involved in creating AI systems.

References:

_https://www.microsoft.com/en-us/research/project/guidelines-for-human-ai-interaction/

https://human-centered.ai/intelligent-user-interfaces-2019/_

Do you have a project 🚀 that you want me to assist you email me🤝😊: wilbertmisingo@gmail.com
Have a question or wanna be the first to know about my posts:-
Follow ✅ me on Twitter/X 𝕏
Follow ✅ me on LinkedIn 💼