Unveiling the Incommunicable: Limits of Conveying Experience

Unveiling the Incommunicable: Limits of Conveying Experience

Communication is a complex interplay of information exchange, often taken for granted in our daily lives. However, delving deeper into the intricacies of how agents interpret and communicate information raises intriguing questions about the limitations of our understanding.

One such thought experiment, famously known as "Mary the color scientist," sheds light on the challenges agents face when communicating about subjective experiences. In this blog post, we will explore the Mary argument and propose a clear explanation for the challenge it describes.

Mary the Color Scientist

The Mary the Color Scientist thought experiment, formulated by philosopher Frank Jackson, poses a compelling inquiry into the limits of human knowledge and the nature of subjective experience. In his paper "Epiphenomenal Qualia" published in 1982, Jackson introduces Mary, a hypothetical scientist who possesses exhaustive knowledge of the physical and neurobiological aspects of color perception but has been confined to a black-and-white environment throughout her life.

Frank Jackson has given the following interpretation of the thought experiment:

  1. Mary (before her release) knows everything physical there is to know about other people.
  2. Mary (before her release) does not know everything there is to know about other people (because she learns something about them on her release).
  3. Therefore, there are truths about other people (and herself) which escape the physicalist story.

The crux of the experiment lies in the moment Mary is released from her monochromatic prison and encounters color for the first time. Jackson's argument revolves around the question of whether Mary, armed with all the scientific knowledge available, truly gains new knowledge or experiences something novel when she sees color. The scenario challenges traditional physicalist views and raises profound questions about the nature of consciousness and the limitations of knowledge acquisition.

The Black Box

Consider a system featuring a black box, a self-contained process shielded from external observation. The black box has exclusive access to both its internal information and incoming data. Within this secluded environment, the black box engages in a series of operations, leveraging the available information.

While the system can produce outputs, these outcomes are not directly perceivable by itself. Instead, they are only acknowledged by the black box itself through subsequent alterations in its input information. In this configuration, the outputs may be accessible to others, yet their true nature and significance are discernible solely to the black box, with the connection between outputs and input information forming an intricate and internalized mode of self-awareness within the system.

In the context of this system, the input information comprises signals with at least two discernible states. These states represent distinguishable conditions or characteristics inherent in the input data. The black box, with its limited access to internal and input information, possesses the capability to discriminate between these states. Whether the input signals manifest as binary code, distinct voltage levels, or any other form with multiple distinguishable states, the black box engages with this information as it navigates its internal processes. The discernibility of these states is crucial for the black box's ability to interpret and respond to variations in the input, shaping the dynamics of its closed-loop operations.

Internal information encompasses a multifaceted repository that includes preexisting knowledge, ongoing thought processes, and established correlations between both input and internal information. The black box, acting as a self-contained entity, utilizes its internal information as a foundation for processing the incoming data. Preexisting knowledge within the black box may consist of learned patterns, rules, or any relevant information obtained prior to or during its operations.

Additionally, the black box engages in thought processes—mental operations or computations that shape its response to the input information. These thought processes are influenced by the internal information and contribute to the system's overall decision-making.

Furthermore, the black box may have established correlations between different facets of information, be it between different inputs or within its internal repository. These correlations serve as a framework for interpreting the significance of incoming signals and contribute to the dynamic nature of the system's responses.

In essence, the internal information acts as the cognitive landscape through which the black box navigates, incorporating preexisting knowledge, ongoing thought processes, and recognized correlations to process and respond to the input information in a meaningful way.

And Then There Were Two

Let us consider how this could look if we had two of these black boxes which wanted to communicate. How would this look?

In a system where two black boxes operate with limited access to only their respective internal and input information, communication between them involves a nuanced interplay of discernible input signals, internal processes, and shared knowledge. Here's a breakdown of what a communication process could look like:

  • Shared Input Signals
    • Both black boxes receive input signals, such as distinct colors, binary patterns, or any signals with multiple discernible states.
    • These input signals serve as the common language through which the black boxes can convey information to each other.
  • Internal Processing
    • Each black box engages in internal processes, leveraging its preexisting knowledge, ongoing thought processes, and established correlations.
    • The internal processes are influenced by the discernible states within the input signals, shaping the responses of each black box.
  • Output Generation
    • The outputs generated by each black box are influenced by its internal processing and the unique interpretations derived from the input signals.
    • While the actual outputs may be accessible to external observers, the nuanced significance and interpretation are known only to the respective black boxes.
  • Feedback Loop
    • The generated outputs serve as input signals for the recipient black box, initiating a feedback loop.
    • The recipient black box interprets the incoming signals based on its internal information and adjusts its internal processes accordingly.
  • Shared Knowledge and Correlations
    • Over time, the black boxes may develop shared knowledge or correlations based on the patterns observed in the input signals and responses.
    • This shared understanding contributes to a more refined and efficient communication process between the two black boxes.

In summary, communication between two black boxes within this constrained system involves a dynamic exchange of discernible input signals, internal processing, and the reciprocal influence of generated outputs. The shared knowledge and correlations developed over time enhance the effectiveness of this communication, establishing a dynamic system of interactions between the black boxes.

The Constraint

In a system characterized by the described constraints, communication becomes bound by the specific limitations governing the exchange of information. The nature of communication within this system is markedly restricted, primarily confined to the unveiling of correlations existing solely within the realms of input information and internal information. The intricacies of these correlations serve as the focal points through which communication unfolds, as the interplay between discernible input signals and the internal processes of each black box establishes the foundation for meaningful exchanges.

Within this confined framework, the nature of communication is notably delimited, relying exclusively on shared correlations embedded within the input information and internal processes. The intricate interplay of discernible states within the input signals and the interconnections within each black box's internal workings establishes the fundamental groundwork for meaningful exchanges.

This constraint can be clearly described. It is solely by the correlation of input signals by which these black boxes communicate. Their communication facilitated only by this shared medium. A key constraint which is often overlooked is that each box does not have access to the other's input information. All that can be established between them is shared correlations within their input information.

A System of Symbols

We could envision such a system to have input information characterized as a set of symbols representing each state of input information. Each symbol being unique in a set per signal. We could concieve of this looking something like a set of characters:

// Agent Alpha - input signals
visualSignalX = [A, B, C]
visualSignalY = [D, E, F]
audioSignalU = [G, H, I]
audioSignalW = [J, K, L]

Given such an example, we could imagine the agent forming correlations to communicate. This would require creating output which altered the agent's own input signals thereby facilitating communication.

// Agent Alpha - before communication
outputSignal = []
audioSignalU = [G, G, G, G, G]

// Agent Alpha - during communication
outputSignal = [?, ...?]
audioSignalU = [G, G, H, H, G]

Using these symbols, patterns could be used to establish correlations between input information. For example, a pattern of G, H, H, G could refer to some specific input symbol like A from the visual input signals. The key constraint here is that it is only a pattern which is used to establish these correlations. The symbols themselves are never transmitted directly.

For this reason, we could describe a second agent which had a completely different set of symbols.

// Agent Bravo - input signals
visualSignalX = [M, N, O]
visualSignalY = [P, Q, R]
audioSignalU = [S, T, U]
audioSignalW = [V, X, Y]

In such a system, communication could still be established using only patterns in input to establish correlations. For example, the output given by Agent Alpha previously could be received as the pattern S, T, T, S by Agent Bravo. Given the constraint of communicating only via correlating patterns, the receipt of information is always restricted to be about input information or preexisting internal information.

It is important to note that regardless of the existence of some form of experience, a system constrained in such a way would be inherently restricted by the method of communication used. Information exchanged between agents could only be about correlations which could be established between input information and/or internal information. This description gives rise to the expectation of these agent's having access to information which is not communicable.

The Argument for Incommunicable Information

In an attempted to crystalize the previous conception of the constraints we work within in the following argument:

  • P1 - If an agent only has access to input information and internal information then correlations can only be established between this accessible information by the agent.

  • P2 - Agents only has access to input information and internal information.

  • C1 - Therefore, an agent can only establish correlation between input-input, input-internal, and internal-internal information.

  • P3 - If communication consists of changing another agents input information then communication of information to an agent is only interpretable via patterns in input information.

  • P4 - Communication is performed via changing of other agent's input information.

  • C2 - Communication to an agent must be interpreted via a pattern in input information.

  • P5 - Correlations between patterns of input and symbols themselves are input-input correlations.

  • C3 - Inaccessible symbols to an agent are incommunicable to an agent.

Given this argument and the example of agents given previously, we can imagine that symbol A is an unknown symbol for Agent Bravo. Agent Alpha would be unable to communicate to Agent Bravo what the symbol A is because the only tool which Agent Alpha has is manipulating input signals which do not contain symbol A.

Rethinking Mary

With the previous argument in mind, we can now reconsider the knowledge argument as conceived in the Mary thought experiment. We can consider all knowledge described in the thought experiment as inherently communicable information. This is because it is information which she learned by being taught by others. The information which is unknown we can consider incommunicable information. In this way, we can reformulate the argument as follows:

  1. Mary (before her release) knows all communicable information about color.
  2. Mary (before her release) does not know any incommunicable information about color.
  3. Therefore, communicable information is not equivelant to incommunicable information.

Seeing as we now can have an understanding of how and why such incommunicable information can exist, it no longer seems like a mystery as to why Mary would have access to such information. There seems to be no reason to attribute extra properties to this information beyond it being incommunicable.

Beyond Arguments For Non-Physicalism

While Frank Jackson's original focus might have been one on non-physicalism, this argument highlights another missing link which seems to be disregarded by some others. Some others may, at times, consider incommunicable information as non-existant as it can not be described. This is a tack in the wrong direction. Simply because it is incommunicable, does not mean it does not exist.

Given the argument provided here, it should be expected that we should have access to such incommunicable information. The existence of this information is the foundation with which we form our communications by using patterns in the shared medium we call reality. There is no need to be afraid of such incommunicable information, nor attribute special properties to it. It seems reasonable to accept that there can be such a thing and there is good reason to believe such a thing exists.

Conclusion

While so many responses to the knowledge argument have been good, I have felt they were always lacking the key point which I have presented here. Specifically, the constraint which we inherently have given the method of communication we use. It may yet be possible that we find some new method of communication. Until that point in time, this seems to be the type of constraint we are working with.

We should not dismiss the existence of incommunicable information. If we are to attribute anything to incommunicable information, it should be for reasons beyond it simply being incommunicable. Incommunicable information should be expected.

Views: -