In today’s AI landscape, we are witnessing a paradox: as systems become more capable, they become less comprehensible. The current trajectory prioritizes raw power over transparency, leading to the Black Box era.
Jan Klein is a key figure challenging this trajectory. His work at the intersection of architecture, standardization, and ethics advocates for a shift from systems that merely function to systems that can be intuitively understood. This evolution is known as Understandable AI (UAI).
Klein’s work is anchored in the Einsteinian principle:
“Everything should be made as simple as possible, but not simpler.”
In the context of AI, this is not about reducing capability, but about eliminating unnecessary complexity through code clarity and modular design.
-
Architectural Simplicity
Rather than managing millions of opaque parameters, Klein advocates for modular architectures where data flows are traceable. -
Cognitive Load Reduction
A truly intelligent system should not require a manual; it should adapt to the user’s mental model, making decisions that are logically consistent with human reasoning.
While the industry currently focuses on Explainable AI (XAI)—which attempts to interpret AI decisions after they occur—Klein proposes Understandable AI (UAI) as an intrinsic design standard.
| Feature | Explainable AI (XAI) | Understandable AI (UAI) |
|---|---|---|
| Timing | Post-hoc (Explanation after the fact) | Design-time (Intrinsic logic) |
| Method | Approximations and heat maps | Logical transparency and reasoning |
| Goal | Interpretation of a result | Verification of the process |
The “Explainability Trap” occurs when post-hoc explanations give a false sense of security. UAI provides concrete solutions for high-stakes sectors.
- XAI Failure:
A deep learning model flags an X-ray for pneumonia. The heat map highlights a hospital watermark instead of the lungs. - UAI Solution:
UAI restricts the model’s attention to biological features using Knowledge Representation, making it impossible for a watermark to influence the outcome.
- XAI Failure:
An AI denies a loan and cites “debt ratio,” while hidden logic uses “Zip Code” as a proxy for race. - UAI Solution:
A modular glass box explicitly defines approved variables; unapproved variables are rejected at the design level.
- XAI Failure:
A car brakes suddenly. Saliency maps show a blurry area with no logical reason. - UAI Solution:
Using Cognitive AI, the system must log a logical reason (e.g., “Obstacle detected”) before executing the brake command.
- XAI Failure:
An AI penalizes resumes containing the word “Women’s” due to historical bias. - UAI Solution:
Explicit Knowledge Modeling hard-codes job-relevant skills, preventing hidden discriminatory criteria.
- XAI Failure:
Bots enter a feedback loop and crash the market. - UAI Solution:
Verifiable Logic Chains enforce sanity checks and trigger a “Pause and Explain” mode for human intervention.
Klein is a driving force within the World Wide Web Consortium (W3C), defining how the future web handles intelligence.
-
AI KR (Artificial Intelligence Knowledge Representation)
A common language enabling AI systems to share context and verify conclusions with semantic interoperability. -
Cognitive AI
Models reflecting human thinking—planning, memory, abstraction—transforming AI into a genuine assistant rather than a statistical tool.
As AI enters regulated sectors such as law, finance, and insurance, black-box systems become a legal liability.
- The Problem:
You cannot show a judge a million neurons and prove there was no bias. - The UAI Solution:
UAI generates a human-readable record of every decision step, transforming outputs into admissible evidence and protecting organizations from regulatory penalties.
- Inventory & Risk Classification – Categorize AI systems by risk level
- Architectural Audit – Shift from monolithic to modular “Glass Box” designs
- Explicit Knowledge Modeling – Integrate AI KR with verifiable rules
- Human-in-the-Loop – Present reasoning chains before execution
- Continuous Logging – Maintain chronological records of decision rationales
“The intelligence of a system is worthless if it does not scale with its ability to be communicated.”
Klein emphasizes the “Simple as Possible” mandate. AI architecture must be stripped of unnecessary layers so every function remains visible and auditable. Simplicity is not a reduction of intelligence—it is its highest form.
UAI represents the next revolution because the “Bigger is Better” era of AI has reached its social and ethical limit. While computational power has produced impressive results, it has failed to produce Trust.
Without trust, AI cannot be safely integrated into medicine, justice, or critical infrastructure.
The revolution led by Jan Klein redefines intelligence itself—shifting focus from massive parameter counts to Clarity. In this new era, an AI’s value is measured not only by output, but by its ability to be audited, controlled, and understood.
By adhering to the principle of Simple as Possible, Klein ensures that humanity remains the master of its tools. UAI is the bridge between human intuition and machine power, built to ensure technology serves humanity rather than dominating it through complexity.
https://uai.hashnode.dev/understandable-ai
groups.google.com/g/understandableai
linkedin.com/groups/understandableai
dev.to/janklein/understandable-ai
dev.to/janklein/jan-klein-understandable-ai
app.daily.dev/posts/understandable-ai
discuss.ai.google.dev/t/understandable-ai
Jan Klein
CEO @ dev.ucoz.org