Researchers using AI have recently determined that there are ‘significant’ differences between the male and female brains. This has profound implications for personalised medicine. Less so for our products. But collaborations between research science and AI could be about to shake up the SAAS industry sooner rather than later.
Buckle up this is going to be a long one
CHARGE syndrome is a complex genetic condition with a wide range of nasty symptoms. The name ‘CHARGE’ is an acronym: Coloboma of the eye, Heart defects, Atresia of the nasal choanae, Restriction of growth and development, Genital abnormalities, and Ear abnormalities. It’s of special interest to scientists studying complexity, because it’s caused by mutations at a specific single locus – the CHD7 gene – which plays a key role in the early stages of development across various organ systems.
The CHD7 gene encodes a protein that is involved in chromatin organisation, which is crucial for regulating gene expression. Unusually, we can identify the trigger (the mutation at CHD7) and observe the downstream effect. Occasionally siblings will both carry the mutation, but one will develop the life-altering symptoms of CHARGE. The other nothing. Why? In the asymptomatic sibling, the mutated gene was ‘switched off’ early. This happens in the body all the time – a gene acquires a methyl cap and expression is dramatically reduced. But why? How? How does one body ‘know’ to eliminate the effect of a rogue gene variant while another doesn’t? We don’t yet know.
The reason the CHD7 mutation is of such interest to scientists studying complexity is that all the effects of CHARGE are caused by single gene. This is actually pretty rare in a complex system like the human body.
A linear ‘flow’ with a range of emergent effects
With CHARGE, we not only see the cause of a range of symptoms, but also all the downstream genes that have their expression affected by how this single gene is regulated.
Epigenetic mechanisms, including DNA methylation, histone modification, and chromatin remodeling, play significant roles in regulating gene expression without altering the DNA sequence. And, more broadly, how the expression of a gene cascades to affect the expression of other genes.
The CHD7 mutation is actually seen as an easy target from the perspective of complexity, because of the unusual simplicity of the ‘cause’ factor, and the profundity of its effect. Most gene interactions are far too complicated for us to identify straightforward cause and effect.
Or a least they have always have been. But AI may be about to change all that.
Could AI help reduce the complexity of complex systems?
Complexity theory explores how complex systems and patterns arise from a multiplicity of relatively simple interactions. It examines how relationships between parts give rise to the collective behaviours of a system and how the system interacts with its environment. Complexity theory focuses on systems that are dynamic, nonlinear, and adaptive, meaning that the whole is greater than the sum of its parts. Examples of complex systems include the Florida Everglades, living cells, cities, the global economy, and the human brain.
Key properties of complex systems (and some of the reasons they’re so hard to predict) include:
Nonlinear interactions: complex systems often throw up unexpected outcomes. Rather than cause A affecting object B, it’s more like the effects of factor D and E combine to alter cause A, leading to an unpredictable A to B effect.
Emergent properties: due in part to these nonlinear interactions, complex systems see emergent properties. In CHARGE syndrome, the length of time the mutated CHD7 gene is fully active doesn’t clearly predict the severity of symptoms experienced by the patient.
Feedback loops and self-organising principles: a complex system will see its variables reacting and altering according to the feedback. For example, a period of heavy rain in an ecosystem might lead to flooding, but as this recedes, soil will be more fertile. The fact that this happened regularly in the Nile Delta in ancient Egypt led to one of the first advanced human civilisations springing up around it. The humans reacted to the ‘feedback’ from the water cycle.
For humans, the way in which all the variables in a complex system will interact is very hard to predict. But this is something AI is very good at. AI can already spot patterns in big data that we can’t. Some have begun to discuss AI hyperreality – the things AI can identify that are simply beyond our perception. Going forward, this will likely be good news for people suffering from complex disorders, from CHARGE to cancers.
Can these principles be applied to our web apps?
Our applications are not as complex as the Florida Everglades. But they are complex. There are a wealth of obscure datapoints. If we had a better idea of these, we might have a better idea of how they might be affecting user experience.
Working on the assumption that there are contextual variables ‘beyond the flow’ could be helpful. And AI could help with that.
Applying the principles of feedback loops and adaptation, borrowed from complexity theory, could help designers create more user-friendly applications:
Implementing adaptive User Interfaces
Feedback loops: collect user interaction data in real-time to understand how users interact with the app. This data serves as feedback to identify patterns, preferences, and pain points. Eg. are some users taking 30 seconds to complete tasks that took test subjects 5 seconds? The AI might indicate ‘who’ is having this issue, which might lead us to ‘why’.
Adaptation: use this feedback to automatically adjust the UI and UX to better match user needs and preferences. For instance, if many users struggle to find a specific feature, the UI could adapt by making that feature more prominent.
Personalisation through learning user behaviour
Feedback loops: track user behaviour over time to glean insights into individual preferences, frequently used features, and ignored ones.
Adaptation: customise the app experience for each user based on this behaviour. This is already happening where, for example, a news app might adapt its homepage to display stories similar to those the user often reads. Not necessarily a good thing!
Enhancing accessibility and usability
Feedback loops: use user feedback and usability testing to identify accessibility issues or areas where users commonly face difficulties.
Adaptation: use AI to identify patterns in the user data – eg. demographics and associated pain points. By using all the data available, and reacting to it quickly, interventions may be as simple as increasing contrast for readability or simplifying navigation at certain points in the flow.
Evolutionary feature development
Feedback loops: implement feature usage tracking and solicit direct user feedback through surveys or feedback buttons to understand which features are valued and which are not.
Adaptation: prioritise development resources to improve popular features, allow unused ones to go extinct, introduce new features, and group existing ones more intuitively based on user requests and behaviour patterns.
Responsive design and content delivery
Feedback loops: analyse device usage patterns and network conditions to understand the context in which your app is most often used.
Adaptation: adjust the layout, resolution, and content delivery based on the user’s device and connection speed to ensure optimal performance and usability across all devices. For example, if we learned that our app was most often used by young parents waiting to collect their kids from school, how might this information affect future updates?
Continuous learning and improvement cycle
Feedback loops: establish mechanisms for continuous collection of user feedback, performance metrics, and error reports.
Adaptation: use this information to inform an automated update and refinement cycle, fixing bugs, enhancing performance, and introducing features aligned with user needs and technological advancements.
Implementation might not be as far off as it seems
We are already not far off being able to:
Automate feedback collection: use analytics tools and embedded feedback mechanisms to gather data without requiring manual intervention.
Dynamic adjustment: develop algorithms and design principles that allow the app to adjust dynamically to user feedback without needing constant manual updates.
User-centric testing: incorporate user testing early and often in the development process to ensure feedback loops are accurately capturing user experience issues. Design surveys to address the ‘whys’ behind big data points to apply real human thinking to solving these problems effectively.
Our ability to implement AI beyond the user flow, to help us address hypotheticals about contextual use could ultimately lead to a more user friendly experience. More about designing within complex system environments here.
1,001 behavioural insights 
Leave a comment