What is a good question?

Questions are transitory requests for a connection between information and range from simple memory retrieval to transformation of understanding. A good question is an Aletheic Probe (resonant connection) because it discloses a reality that was hidden rather than Veritas (only looking for truth). Humans would describe this disclosure as insight where the new understanding resonates, goes beyond a fact and feels right. AI maps data to a lower dimensional manifold and grokked as there is a sudden phase transition. 

Usually, a human who finds meaning in a pattern will experience resonance, the pattern will ring true with their internal prediction. The feeling of surprise when our active inference fails to predict is called cognitive dissonance. As a disability analyst my focus is understanding the emotion responses to this dissonance. People may see this feeling as an opportunity to learn and or put up a barrier. A good question should catalyse the process of finding the answer. 

A good question will evoke cognitive dissonance which provides the discomfort necessary for change. In AI terms this is a high energy high loss state that can be used for gradient descent towards an answer. Humans use collaboration to increase the noise so that a resonant pattern of meaning appears. Paradoxically low noise does not increase the signal and makes it easier to get stuck in local minima (such as group think). The learning environment (e.g. with AI or university) will influence how easy it is to see the resonant patterns.

Helping people to grok

Humans (and AI) need more than one glimpse of pattern to be able to see it. A question is transitory and will light up the area of the question once. There must be some form of noise to ensure that the pattern is lit up repeatedly. Ideally the question should have three elements of pattern, meaning, resonance that each pluck at the strings of the understanding. Some ambiguity in the question so that the person must re-read the question is also helpful. To create a latent space the question must not threaten the person’s sense of self. 

Education is a statistical process, many people are offered many ideas and overall some will develop higher level understanding in some areas. There will be a proportion of students who are mainly rote learners and have limited understanding how to generalise the information. Others will struggle to memorise the facts and derive from first principles. Good questions can both improve retention of facts (with hooks) and understanding of principles (grokking) but not attempt to teach everything to everyone.

Transcendence 

The word transcendence was suggested to me by AI and the sense of spiritual experience describes the feeling of resonance. Recognising a pattern has a pleasant feeling, seeing the meaning within that pattern is exciting but the resonance goes further. Scientists and philosophers may not describe this resonance as transcendence but often spend their whole lives trying to connect with something greater than themselves or as Hartmut Rosa puts it ‘vertical resonance’.

Often described as a search for truth or enlightenment it suggests that our brains have an emergent property. We can detect the architecture of reality itself and our constructs (or manifolds) have testable consequences. Mathematical representations of physics are an example of this phenomenon and its limitations. We know that current theories are incomplete but despite this they are useful. Transcendence may only be our reaction to shadows of reality but still represent a major advance in human understanding. 

Collaboration

A good question will provoke discussion and debate, it will have space for different perspectives and points of view. By applying different people’s assumptions of the way to answer there will be an information bottleneck. Each speaker will focus on only some of the information in the area causing the question to become simplified. As the group between them holds the whole information each listener will be able to compare their view with that of the speaker slowly building a holistic understanding. 

The group has another emergent property as any individual could identify resonance but together they can amplify that resonance. If there is a general solution then many people in the group will sense the resonance at the same time. This property ensures that transcendence is a phenomenon of the group rather than individual making it more robust. If the resonance is seen in many viewpoints it is more likely that they are finding something real. 

Conclusions

To determine whether a question is ‘good’ requires an understanding of both the respondent and the purpose of the question. A respondent who does not does not feel safe to give their point of view will get less from a question. A question which engages the person’s pattern recognition, meaning detection and even resonance will be more effective than information retrieval. Grokking requires emotional as well as logical engagement to be effective. 

Emotional intelligence depends upon the ability to ask meaningful questions that resonate with human experience. Writing a good question is half the journey towards better understanding. For AI to ask good questions it needs to have information on the person’s emotional state, who else is there, whether they can see the pattern and meaning. The correct timing of the question will also influence how effective it is. 

In this article I have assumed that a ‘good’ question is an Aletheic Probe as this creates a resonance. I have also considered asking a question as part of an interaction with a respondent or a group. This approach is based on the idea that prediction error can create a resonance (and possibly a phase transition) in the reader. There are many other ways of looking at questions which will offer different resonances which may now be clearer.

May 2026