You can find part-1 in case you missed it.
Human Machine Diagram
Here, we should look at the entire picture in a logical way. When we say “human” in the first circle, we refer to the person who writes the algorithms from his own viewpoint. Then, the machine learns the algorithms to serve the other “human”, i.e. the user. The person who develops such a product should be aware of all dimensions, and this is the important issue for the user, whose behavior determines the way of using such a product.
Moreover, the ML or the AI can’t recognize the problem as a problem like the case of Google Mail.
Google mail attach
Currently, Google Mail notifies you in case you forget to attach a file, if your mail text has a word referring to attachment. However, the AI is already used, but it fails to solve this problem.
I think that the Data Science realized that people forget the attachment, then this problem was learnt (learning), and an action was taken by notifying the user about the attachment.
So, the role of the person who develops the AI is to be aware of the objective behind the developed product. When we are about to develop an AI product, we should ask ourselves the question in a different way,
Instead of asking “Can we use AI to …….. ?”
We should start exploring human-centered AI solutions by asking:
How might we solve …….. in a unique way?
• • •
Lagorta as a practical example
The below module in the project, simply, was about building a campaign by using AI. We had to collect and sort data to use it in helping users. We wanted to have all tasks done by the Engine as soon as the objective, goal, and budgets are written.
This is one of the long-term projects we have worked on. An AI-driven autonomous growth engine and we support in the era of experience and design. in addition, humanist AI.
Lagorta — AI-driven autonomous growth engine
We had to place two inputs to achieve this goal. After UX searching on the issue, we found that the users will not do those tasks!! So that this huge and the strong AI engine will not have that value!
After iterations and using human-centered AI activities
We tried to make solutions that push the users to fill the required inputs without the need to make anything as a mandatory field! therefore the design, human-centered AI helps both business and technology achieve their goals.
• • •
This is to say that we have to have a comprehensive look. It is not enough to have a strong Engine, it is important to be aware of the users’ behaviors and their context and needs of the Engine.
So, we need to accept human behavior the way it is, not the way we wish it would be.
Accept Human Behavior
Accept human behavior as it is, and we don’t have to change human behavior to achieve our goal.
For example, if you have a door and you don’t know how to open it; should I pull, push it?
The answer maybe is to put signs on the door to tell the person what to do to open it. Well! If we put signs and the person opens the door correctly, how long did it take him to think about the proper way of opening the door? We provided the person will many instructions that made him disturbed.
The same happens with people when they don’t find a product that solves their problems fast and effectively. From our point of view, the door we talked about is functional; it has a handle, it can be opened and closed, it has a sign on it and it has all options, but you need to change the user’s behavior to make him accept this door.
However, if we accept the user’s behavior and we work on this behavior, we will get a better idea. This is the model of Don Norman, the father of UX. He remove the handle on the push side to convince the user in one certain direction.
If we take the previous diagram on a larger scale, we will see that we have humans, products, and a business. The business is the core of investment and the investment aims at raising profits. We also have a machine and of course, we have an ecosystem, in terms that each product depends on its preceding products or application, or simply depends on the AI Engine updates of the algorithms according to the needs coming from the Data Science part.
Let’s have an example from the world and the context in general.
Suppose that we have a person who urgently needs medicine from a pharmacy, and there is a robot that will bring the medicine for that person. The robot went to the pharmacy and found a long queue. So, the robot decided to bring the medicine himself and go back to the person who needs it without paying for the medicine.
Is it the proper behavior? The answer is no. Because, every time the pharmacy is crowded, the robot will do the same action even if the person is not in bad need for the medicine since this is the main metric of the robot is to get the medicine quick.
Also, there may be some old people in the queue who can be disturbed by the robot’s behavior. So, it is very important to fully understand the context in which the AI will operate.
I’m trying to focus on the digital products as I want you to focus on the entire concept regardless the way of applying the concept.
The digital product will be a part of the full product, so it is important to be aware about this and to have these skills.
• • •
The Design of AI has its foundation. So, we need to set rules and ethics for AI. The University of Stanford, the EU, and Australia are now working on setting rules and ethics for AI. I hope we will do the same in MENA.
To have useful AI, we should have a clear goal. This allows us to ask ourselves the question in a different way. Instead of focuses on what we do with AI, We should ask ourselves how to solve a problem in an innovative way and how AI can help us do this in a unique way. This point of view is important to achieve the product’s purpose.
2- The Value:
AI isn’t a dazzling technology. it should extend the value for the user that will use the product. also for the business.
This is the most important part. If AI is not bound by ethics, many problems will appear. For example, Amazon’s recruitment engine was biased for males for certain reasons. So, we should care about ethics to avoid disasters.
No goal can be achieved without trust. The users’ trust is vital for buying a product. For example, I deactivate the privacy properties on Facebook because I don’t trust Facebook when it comes to my private data. This leads to a problem in data collection.
• • •
We need to be aware of some aspects when we think about AI:
1- Stay Human-Centered
We should be human-centered. We should focus more on the users than the technology. Innovation shouldn’t take us away from the users’ needs.
2- AI systems should help humans.
For example, Netflix built strong trust with the users who developed a positive impression about the Netflix content.
Also, Netflix is user-oriented as it tailored the movie thumbs according to the users’ behaviors.
3- We should make fake tests to reach a strong result.
For example, the BOT couldn’t differentiate between users’ orders editing and new orders.
4- Understanding Humans
Humans communicate with each other in a certain way, and when a machine communicates with humans it lacks human sense, which is necessary for effective interaction. So, we need AI that is able to understand humans.
• • •
Finally, We should focus on the notion that we need to reach a new balance between humans and machines.
Humans need certain aspects like ethics, while machines focus on digits and the correct action to be taken. The best solution is the balance between humans and machines.
• • •
How — Human-centered AI
Based on IBM, The framework is a set of activities, tools, and principles that enable teams to design thoughtful, human-centered artificial intelligence solutions using Enterprise Design Thinking.
The AI Essentials Framework is a specific grouping of activities to work through to align your team on strategy for an AI experience. There are five focal areas in the framework:
IBM – Team essentials for AI
- Intent: Align on the business and user intent(s) for your solution.
- Data: Document the data you could use to make your idea a reality.
- Understanding: Determine what you will need to teach your AI.
- Reasoning: Bring your ideas down to earth.
- Knowledge: Brainstorm the direct and indirect effects of your AI.
you can follow IBM or Google AI+People frameworks to start your first project very soon!
Before ending my talk, I’d like to thank you. I hope my talk was useful and meaningful. Your feedback will tell me about it.
I hope that Google will support this issue or at least tackle it and I hope to participate in such dialogues.
Thank you, and I hope to meet you again soon.
[Arabic] Designing Human-Centered AI – Tarek Elsawaf
Thanks for reading!
You can find part-1 in case you missed it.