AI experience manager for the network of intelligent ecosystems
To create intelligent ecosystems, the exchange of data is an essential element. Personalized services and communication in a context won’t happen unless the right type of information is sensed, translated and communicated. With the growing consciousness around data privacy, the solutions for data management are on the rise. The lack of transparency behind the algorithmic decision making is another aspect that utilizes the need for the users to be able to have insights into how they are scored, judged and what goes on behind the algorithmic curtains. I would like to present a concept for managing the identity for the age of distributed internet. (Please note that this identity solution does not need to apply to humans exclusively and could be used as a direction for building intelligent ecosystem identities)
The market these days provides us with many identity solutions being established but none of them unifies the user with the ecosystems emerging. They don’t bond the user with the real value of the services which require real-time ongoing data feed and doesn’t allow the user to control what organizations and consequently the intelligent systems “think” about us. They don’t allow the users to be in a constant conversation with the ecosystems surrounding us. People just like organizations need a solid governance system in order to be a key player in the emerging future. I would like to pinpoint the most crucial aspects of the AI hub model that I have created. I will also talk about how that communication via the AI hub and the emerging systems should look like.
Key elements
Personal cloud
In order to make the automatic systems correspond to the individual user context, the data needs to be owned by the users who can grant permission for various levels of transparency with a specific company. Unfortunately, the current privacy ecosystem is not scalable and efficient. Having a digital identity governed from one access point that manages the data that we produce seems to be a solution. A person, holding an identity, owns the value they create. Users are able to track the use of their data and they can trade it with the third parties who can, in the end, expand and enhance their business models. The user at this point is capable of sharing meaningful information to the third parties in order to receive fully personalised high-quality services. Different levels of engagement can be established too depending on the needs of the user. The levels of engagement carry along the amount of sensitive information that would be required in order to deploy the engagement and provide a service. Whatever services the user is using the output recorded from these interactions should remain at the core of the identity enriching the profile and serve as a direction for further interactions.
On-device AI
“We want to empower people with new AI-driven abilities. With our advances in on-device AI, we can develop new, helpful experiences that run right on your phone and are fast, efficient, and private to you” — Google
There has been a trend of deploying AI solutions towards the edge devices. It has many advantages since it is faster, more secure, saves power and doesn’t require much network bandwidth. Running on-device AI would centralize the operations run on top of the user identity and allow for more sophisticated communication between the identity and the identity of the surrounding ecosystems. On-device AI would be a governing point for our identity and would connect and collaborate with other AI/ML solutions, working on our behalf without revealing sensitive information about us. The data that would be sent back to the organization providing us with services wouldn't contain personal data but rather the information necessary to improve and enhance their existing solutions. The user would be given a set of tools to work together with the on-device AI to constantly keep improving his/her profile so that the services provided are more accurate and personalized over time.
Distributed technology
In 2017 Google introduced Federated Learning (FL), “a specific category of distributed machine learning approaches which trains machine learning models using decentralized data residing on end devices such as mobile phones.”
when it won’t negatively impact your experience. Link to the source
The next big thing is not the thing but how the entities within distributed systems collaborate and enrich each other with valuable information about other ecosystems and human users. Distributed technology is a solution that shares the information, resources and its capabilities with other participants in the network without storing it centrally. It provides users with a single image of the system. It is easily scalable and modifiable. In terms of identity management, it is undoubtedly a “must” solution in order to create coherent solutions that work in tandem with each other. The distributed solutions are necessary to present information to the users it in a contextual way — the right time of the day, type of activity, events taking place or even current mental health; to create meaningful connections in the intelligent ecosystems allowing for human mobility- intellectual, emotional and physical.
AI HUB
The identity management app should be structured in a way where the identity is at the core of it and it branches into other micro identities manageable from the same point. By identity, I mean a set of legal and non-legal verifications that will qualify the user to be an eligible and recognizable entity within the intelligent ecosystem. The same way the micro identities should work but in addition to that, there would way more purpose-driven and of specific characteristics. Therefore, I have introduced a scoring system for each of them. Each user’s identity should possess a revenue system which would consist of two parts. A data wallet and income wallet. The sources might vary depending on what entities the user is interacting with. Additionally, I have added an AI profile which acts as a bridge between us and our automation. It can help to adjust the technology to our personality. it could guard us if some misalignments occurring against our set of values. It is like choosing how we want to be seen, what do we want to see instead of passively being scored without acknowledgement and a power to change anything that we receive in form of content or the service.
Biometric identification
Identity system would be accessible by the biometric identification. Biometrics is an automated method of recognizing a person based on psychological or behavioural characteristics. This way of authentication, first of all, is very secure and secondly, it could be used to gather data about our emotional states and patterns in brain activity that could be valuable to improving/ adapting the solutions running on our devices. Instead of asking the user about our emotional states and health conditions which could be rather inaccurate the biometrics don’t rely on human judgement but purely on the technology. By looking at the sequence of images the user not only unlocks the device but also updates his/her AI profile by sending data about its biometrics. By looking at the sequence of images the system is able to record the brain activity which is generated as a response to each image. Brain waves carry information about the emotional states. Like for example, alpha waves are connected to a relaxed, tranquil mental state. Not only the biometrics can inform the services running about our health conditions in a broader picture but they can also inform real-time running solutions such as adjusting the experiences. For example, in the paper entitled Using Pupil Diameter to Measure Cognitive Load by Georgios Minadakis and Katrin Lohan the study has been conducted about the possibility of measuring the cognitive load by eye-tracking solutions. The cognitive load feedback can further inform the user interfaces that can be adjusted to the user’s needs. The high cognitive load could potentially reduce the complexity of the required tasks, change the tone of the voice of content provided, alter the visual side of the communication and make the decisions about the type of tasks to ask the user or even abandon certain type of actions and instead focus on improving the wellbeing of the user.
Profiled identities
The identity system should allow people to have different identities attached to different major aspects of their life like for example health. Health profile can be very different from a financial one. There could be different motivations behind each of them. This is the reason why I have identified separate sections for them within the AI hub. Within one profile the user would be given a score/ reputation for each identity that is supposed to help you see how the companies that the subject is working with are profiling him/ her but also what services are used that influence the profile. Each of the identities receives a set of automated tasks that the subject is able to perform just within one click such as sign up for a health check or perform a diagnosis at a designated medical location. The suggestions are preselected and filtered through the AI profile [See Fig.16 and Fig.17]. The identities should be pulled up in a specific context to leverage the best sort of experiences. The micro identities would always work in tandem with the main identity. The best aspect of having customized content and service pulled in front of your eyes is that it limits the user’s time of searching for the right type of information and prioritizing the meaningful set of actions to be taken without intrusive notifications coming all at once. Notifications are actionable in AI Hub solution and they are always tightened with an improvement of the system by using a wide range of the user inputs.
If the subject would like to become a better professional, the career identity would show which conferences I should go to, which people I should connect to, what are the courses I should take etc. It should anticipate my actions and make them happen within a click of a button. As an example, there is a conference coming up. The career identity suggested it. The subject clicks a button “Attend” and the chain of actions follow. I receive a notification on my phone that a conference is happening on this date and this address. The automated system should book these conference tickets for the user and put it on the calendar. It should inform the user about what is happening at this conference and save this information in a preferable place. It should anticipate that the user might need a place to stay and if that is the case, book a hotel for the user and save all the booking details in a preferable place.
Companies might put special offers and create small visual ads on the home page. The choice of who is able to put these ads is carefully selected so it matches the user’s profile [See Fig.4].
Building trust by being authentic and showing competencies
It might happen that the algorithm doesn’t have enough data to support the autonomous decision for choosing one service over the other or to make a suggestion [See Fig.5]. In that case, it is going to ask the user to make the decision. But it will not leave the user without any support. The AI system is going to show its competencies and provides you with a sufficient amount of information so that you can easily make that decision [See Fig.7]. It will research and analyze the options and give the user options to make a further decision. The system will provide you with the information on both services suggested and make a comparison based on the choices made by other people with similar profiles and the analysis based on the user’s profile. The user has also an option to give the freedom of a choice to the system so that AI system can make the decision. Instead of being 100% certain the system will ask for the input and make these good decisions together with the user. It also helps the user to understand how the algorithm works and also know its options too. And most importantly it builds trust and belief that the system is competent enough to provide relevant information helping the user to make a good decision and follows a logic that the user posses and uses in everyday life.
Another case would be when the algorithm is not able to make a decision or is missing a piece of data in order to make a decision it will ask the user for further input [See Fig.6] so that the system can make a decision after all. That way the AI hub learns about the user over time by interacting with him/ her. It creates a form of trust and keeps the user in control/ at the core of his/ her identity. It is very hard to assume that the identity solution will know the user from the beginning. Trust can only be built over time.
Inverting reputation
Ranking systems are widely applied mechanisms used to shape the desired behaviour either in human or a system. They allow agents of the system to evaluate the qualities, experiences, and behaviours of others to distinguish between those who benefit the system and those who don’t. One of the known methods is the Social Credit System. It is a program run by some governmental agencies with the ambition to collect information available online about a country’s companies and citizens in a single place; the data is then used to assign each of the individuals a score based on their demographics, commercial, online and offline behaviour and social network. Are these ranking systems bringing positive outcomes in the long run? How could we learn from this example and make sure that we are building systems that use scoring algorithms for value-adding reasons avoiding setting norms? Scoring systems could be used as assessment tools to assume the appropriate characteristics of the intelligent ecosystems so the right experiences can be formed around them. The forming ecosystems could have their performances measured too in order to establish their reputations and how they match with other ecosystems in the network. Designers would need to develop the experience metrics for each of the ecosystems against which the measurements could be taken. And that means the experiences happening within these ecosystems would need to be designed to be not only qualitative but also quantitative.
Scoring is an inevitable technique in data-driven societies but what if the same system would help people to assess the trustworthiness of the service providing companies? What if people could follow the gamification principles of having a better score to become better versions of themselves and to possess better financial or health reputation? Scoring could reveal the weak points in society as well. For instance, let’s think of a data set that has information about people’s debt, where people with debt would typically receive a lower score. How could we use the information pulled from the profiles of those people to build systems that help to prevent people from having debt in the first place? Instead of teaching AI to solve problems why can’t we teach it to prevent them? Using this principle, the focus to machines smarter would instead shift to let the machine enhance the human capabilities and communicate with other ecosystems to search for the roots of the arising inequalities, faulty behavioural aspects of human and the system and finally, adjust the governing models.
The system provided in AI hub is transparent and what you see in your profile is what health organisations see as well. Health identity and any other identity has a score. The way the score is generated is to help you build up a reputation and to be able to see how the health ecosystem is assessing you. The user is able to build a reputation over time by signing up for services that are recommended to him/ her.
If there is any suggestion that the user doesn’t agree with or thinks it doesn’t match your profile the user should be able to mark it. And the readjustment should be made by the algorithm responsible for the health identity.
In addition to participating in the services, the user should also be able to store all the documents in one place [Fig.10]. That should help him to get more personalized suggestions. And these documents should act like smart contracts. Meaning they should be a valuable piece of code that can be understood by any other intelligent ecosystem and be used to create another connection within these ecosystems to provide customized services. Providing documents to the identity adds up to the score as well as they help to build a profile. Having a higher score means you are a valuable entity in the network and other systems can trust you more and engage on a more personal level.
Sharing credentials
Whenever an organization request any data from the user, the user will get a notification and will be able to share his/ her credentials. He/ She will be notified which data will be shared and he can pick what he wants to share and what not. If there is any data that our AI system doesn’t recommend us to share the user will be notified [Fig.11].
The currency of the automation
Loomia is taking the first steps towards the privatization of data and building a platform where users can securely store their data, manage it the way they want – either keep it or sell it. The Loomia Tile gathers information about the use of the clothes, such as frequency and length of use. Users can decide whether they want to share this information with merchandise in exchange for blockchain rewards. Snips provides decentralized voice assistant technology which doesn’t send personal data to the cloud. Instead, the Snips system offers reimbursing consumers in exchange for contributing their encrypted data to its blockchain network.
Data is an asset with value and this value changes over time depending on what happens to it. In AI hub I have created a system that if the company decides to use user’s data for a project that will bring revenue, a token should be released to the user. The token should act as a stock of the project’s revenue so it can become more valuable over time if the project succeeds. In that way, the users get rewarded for his or her contribution in an ethical way [Fig.14].
The wallet section in the AI hub [Fig.12] should be divided into two sections. One section will display rewards for sharing data with different organizations and their projects. The user is able to track how much he is earning over time by sharing data as well as how much he makes per specific project. How his data is used and how it is monetized.
The second part is would be about the user’s income coming from different sources combined.
Some of the tasks performed daily will be automated such as paying bills or paying for the services that we frequently do or use. In this case [Fig. 15], the user charged his EV and after completing it he just gets notified about how much he was charged. No interaction with the device was further required.
AI profile
AI profile came from the idea that people should be able to be in control of their data but also with what algorithms think about us. As we go further with the automation, what type of services do we get will derive from the analogy of algorithmic decision making and how we are profiled. So I would like to introduce the AI profile which allows the user to be able to see how the user is perceived as an entity within the intelligent ecosystem. The entity with motivations and personality but also how he/she is perceived from a health profile perspective or financial one. This method would remove bias and include each person’s needs and create a more stable self-improving autonomous system. The sophistication of that system should increase over time as the services improve and therefore more tailored options can appear in this AI profile. The user can go more into detail and adjust for example what are his motivations.
Ecosystem identities
Just like the users would possess the centre for managing their on-device AI identity, the ecosystems should be manageable too. The intelligent ecosystem should possess its own policies, motivations, value proposition, core structure, revenue streams, key activities perhaps personality too. The people managing the ecosystem should construct these systems for the user input so their experiences and interfaces can be adjusted according to its characteristics. We could set our priorities, channels of interaction and the frequency of the contact. So one hand it would be automatically adjusted to our personal identity but it would also contain a set of settings that allow the user to remain in control of. Each specific situation or a service provided by the ecosystem should allow for the human feedback to leverage the cognition capabilities of acquiring knowledge or experiencing specific service.
References:
Using Pupil Diameter to Measure Cognitive Load by Georgios Minadakis and Katrin Lohan, 2018
Reimagining the Goals and Methods of UX for ML/AI by Philip van Allen, 2017
Federated Learning: The Future of Distributed Machine Learning by Mi Zhang, 2019
If you are interested in learning more about my approach to designing an identity for the distributed autonomous economy you can read my other article Designing a decentralized profile dApp published by UX Collective.
Don’t hesitate to get in touch via email, or👇
- 👩💻 Check my shots on dribbble
- 🔧 Study my use cases on Behance
- 📖 Read more here on Medium
- 🐦 Follow me on Twitter
- 👍 Like me on facebook
