Quacking the Code: AI-Augmented Rubber Duck Debugging and the Future of Cognitive Enhancement in Software Development

by Dr. Jerry A. Smith

Background: Me and My Rubber Ducky

Picture this: It's 2 AM, the office is empty save for the soft glow of my monitor, and I'm locked in an epic battle with a particularly stubborn bit of complex code. My trusty companion in this late-night coding crusade? A small, yellow rubber duck perched on my desk, its perpetual smile a beacon of hope in my sea of curly braces and semicolons.

"Listen here, Michael (my ducky’s name)," I mutter, my caffeine-addled brain searching for clarity. "We've got a nested loop that's behaving like it's stuck in a temporal vortex. Walk me through this..."

As I ramble on to my plastic confidant, something magical happens. Explaining my code aloud, line by line, to this inanimate object begins to untangle the knots in my logic. Suddenly, it hits me mid-sentence – that glorious "Aha!" moment that every developer lives for. The bug reveals itself as if Michael’s unwavering gaze had spotlighted it all along.

Little did I know, my late-night tête-à-tête with Michael was more than a quirky developer habit. It was a gateway to understanding the intricate dance between cognition, verbalization, and problem-solving. As AI enters the scene, our rubber ducky debugging sessions are about to get an upgrade that even Michael couldn't have seen coming.

So, grab your favorite rubber duck (or debug buddy of choice). Let's dive into the fascinating world where software development meets cognitive science, and AI is ready to join our debugging party. Who knows? By the end of this journey, we might find that our little rubber friends have some digital competition in the problem-solving department.

Abstract

This article explores the novel intersection of artificial intelligence (AI) coding assistants, such as GitHub Copilot, with established metacognitive techniques in software development, mainly focusing on the synergistic enhancement of rubber duck debugging. We propose a framework for understanding these AI tools as external manifestations of subconscious cognitive processes, drawing parallels with the psychological concept of tulpas. Through an in-depth analysis of the neurological and cognitive mechanisms at play, we examine how this integration potentially amplifies problem-solving capabilities, accelerates insight generation, and reshapes the landscape of software development practices. This study aims to provide a comprehensive understanding of the emerging paradigm of AI-augmented cognition in programming, its implications for developer productivity and creativity, and the broader impact on the future of human-AI collaboration in complex cognitive tasks.

1. Introduction

The software development landscape is undergoing a profound transformation with the advent of AI-powered coding assistants. These sophisticated tools, exemplified by GitHub Copilot (Chen et al., 2021), leverage large language models trained on vast code repositories to generate contextually relevant suggestions in real time. Concurrently, traditional metacognitive techniques such as rubber duck debugging continue to be valued for their ability to externalize thought processes and facilitate problem-solving (Hunt & Thomas, 1999).

This article posits that integrating AI coding assistants with established debugging techniques, particularly rubber duck debugging, creates a synergistic effect that potentially amplifies subconscious cognitive processes and enhances problem-solving capabilities in software development. By conceptualizing these AI tools as technological analogs to externalized cognitive constructs, we propose a novel framework for understanding and optimizing human-AI collaboration in coding tasks.

The objectives of this article are threefold:

1. To examine the cognitive and neurological mechanisms underlying rubber duck debugging and its interaction with AI coding assistants.

2. To propose and analyze a model of AI-augmented rubber duck debugging.

3. To explore the implications of this synergistic approach for cognitive enhancement in software development and beyond.

2. Background and Theoretical Framework

Before we dive into the intricate interplay between AI and rubber duck debugging, it's crucial to establish a solid foundation of the key concepts at play. This section will explore the cognitive science behind rubber duck debugging, the fascinating world of tulpas and externalized cognition, and the cutting-edge capabilities of AI coding assistants. By understanding these individual components, we can better appreciate the revolutionary potential of their integration in software development practices.

2.1 Rubber Duck Debugging: A Metacognitive Approach

Rubber duck debugging, a term coined by Andrew Hunt and David Thomas in their seminal work "The Pragmatic Programmer" (1999), refers to a method where a programmer explains their code line-by-line to an inanimate object, typically a rubber duck. This technique is grounded in metacognition, the awareness and understanding of one's own thought processes (Flavell, 1979).

The efficacy of rubber duck debugging can be attributed to several cognitive mechanisms:

1. Verbalization: Explaining code aloud engages both Broca's area, responsible for speech production, and Wernicke's area, involved in language comprehension (Flinker et al., 2015). This dual activation may facilitate a more comprehensive processing of the code's logic.

2. Perspective Shift: By explaining the code to an imagined naive listener (the rubber duck), programmers are forced to adopt an external perspective, potentially highlighting assumptions or logical flaws not apparent from their original viewpoint (Miyake & Norman, 1979).

3. Working Memory Offloading: Verbalization may serve as a form of cognitive offloading, freeing up working memory resources for problem-solving (Risko & Gilbert, 2016).

2.2 Tulpas and Externalized Cognition

From Tibetan mysticism, Tulpas refers to mental constructs that allegedly develop autonomy through intense focus and imagination (Veissière, 2016). While controversial in mainstream psychology, the tulpa phenomenon provides an intriguing framework for understanding externalized cognitive processes.

Key aspects of tulpas relevant to our discussion include:

1. Perceived Autonomy: Practitioners report tulpas developing independent thoughts and behaviors, analogous to how AI coding assistants can generate unexpected or novel code suggestions.

2. Interactive Dialogue: The reported internal dialogues with tulpas resemble the back-and-forth interaction between programmers and AI coding assistants.

3. Knowledge Access: Tulpas are sometimes perceived as having access to knowledge or insights beyond the conscious awareness of their creators, paralleling the vast knowledge base of AI coding assistants.

2.3 AI Coding Assistants: Technological Manifestation of Externalized Cognition

AI-powered coding tools like GitHub Copilot represent a technological manifestation of externalized cognitive processes. These tools utilize transformer-based language models, such as OpenAI's Codex, trained on vast code repositories to generate contextually relevant code suggestions (Chen et al., 2021).

Key features of AI coding assistants include:

1. Context-Aware Suggestions: The ability to understand the surrounding code context and provide relevant completions.

2. Multi-Language Support: Proficiency across various programming languages and frameworks.

3. Real-Time Interaction: Immediate generation of code suggestions as the programmer types.

4. Pattern Recognition: Identification of complex coding patterns and best practices from its training data.

3. The Synergy of Subconscious Processes and AI

We propose that AI coding assistants can be conceptualized as technological analogs to externalized subconscious processes, like how tulpas are perceived in certain cultures. This framework suggests several key synergies that emerge from the interaction between human cognition and AI systems in software development.

3.1 Enhanced Pattern Recognition

The subconscious excels at pattern recognition, often processing information and identifying patterns before conscious awareness (Kihlstrom, 1987). This ability is crucial in programming, where recognizing code structures, algorithmic patterns, and design paradigms is essential for efficient problem-solving.

AI tools like GitHub Copilot, trained on vast code repositories, can significantly augment this innate human capability. These systems can identify patterns across a much broader range of coding contexts than any individual developer could experience in their career (Chen et al., 2021). This expanded pattern recognition capability manifests in several ways:

1.        Cross-language pattern identification: AI systems can recognize similar patterns across different programming languages, potentially helping developers apply concepts from one language to another more easily.

2.        Best practice suggestion: By analyzing millions of code samples, AI can suggest industry best practices and design patterns that may take time to get to the developer.

3.        Anti-pattern detection: AI can identify potential anti-patterns or code smells, helping developers avoid common pitfalls.

4.        Contextual pattern application: AI can suggest how common patterns might be adapted to the specific context of the current coding task.

This synergy allows developers to recognize and apply complex coding patterns more efficiently and effectively, potentially leading to higher quality code and more elegant solutions.

3.2 Accelerated Problem Solving

When combined, the rapid processing capabilities of the subconscious mind and AI systems can lead to accelerated problem-solving. The subconscious mind often works on problems in the background, leading to sudden insights or "aha" moments (Kounios & Beeman, 2009). These moments of insight are characterized by a sudden restructuring of the problem space, often leading to novel solutions.

AI coding assistants can complement and potentially accelerate this process in several ways:

1.        Rapid solution generation: While the subconscious mind is processing the problem, AI can quickly generate and present multiple solution possibilities, providing a broader set of options to consider.

2.        Alternative perspectives: AI suggestions may present the problem from different angles, potentially triggering new associations in the developer's mind.

3.        Steppingstone solutions: Even if the AI-generated solutions are imperfect, they may serve as steppingstones, providing partial insights that guide the developer towards a complete solution.

4.        Constraint relaxation: AI suggestions might implicitly challenge the developer's assumptions, helping to relax self-imposed constraints that may be hindering problem-solving.

This synergistic process can potentially trigger or accelerate insight moments, leading to faster problem resolution and more innovative solutions.

3.3 Extended Knowledge Access

While the human subconscious draws on personal experience and implicit learning accumulated over a developer's career, AI assistants access a vast, collective knowledge base derived from millions of lines of code written by developers worldwide. This combination allows developers to leverage both deep personal insights and broad, community-derived knowledge (Boden, 2004; Chen et al., 2021).

This extended knowledge access manifests in several ways:

1.        Exposure to diverse coding styles: AI can introduce developers to coding styles and approaches they might not have encountered in their personal experience.

2.        Domain-specific knowledge: For specialized domains, AI can provide relevant code snippets or patterns that the developer might not be familiar with.

3.        Up-to-date practices: As AI models can be regularly updated, they can introduce developers to the latest coding practices and library usage.

4.        Rare use cases: AI can suggest solutions for rare or edge cases that the developer might not have encountered before.

The AI can thus fill knowledge gaps or suggest alternative approaches that may not be immediately apparent to the developer, effectively extending their cognitive reach beyond their personal experience.

3.4 Dynamic Feedback Loop

The interaction between the developer's subconscious processes, verbalization (in rubber duck debugging), and AI suggestions creates a dynamic feedback loop. Each component informs and enhances the others, creating a synergistic problem-solving environment:

1.        Verbalization stimulates subconscious processing: As the developer explains their code or problem to the "rubber duck," this verbalization can trigger subconscious processing, potentially leading to new insights.

2.        Subconscious guides AI interaction: The developer's subconscious understanding of the problem influences how they interact with the AI, guiding the types of suggestions they seek or how they phrase their queries.

3.        AI suggestions influence verbalization: The code or solutions suggested by the AI can shape how the developer further verbalizes the problem, potentially leading to more precise or different explanations.

4.        Verbalization refines AI suggestions: As the developer continues to verbalize their thoughts, this can provide more context to the AI, potentially resulting in more relevant or refined suggestions.

5.        AI suggestions trigger subconscious connections: The suggestions provided by the AI might trigger new associations or ideas in the developer's subconscious, leading to novel insights.

This dynamic feedback loop creates a rich, interactive problem-solving environment that leverages the strengths of human cognition, verbalization techniques, and AI capabilities. It leads to a more thorough exploration of the problem space and more innovative solutions than either human-only or AI-only approaches could achieve independently.

4. AI-Augmented Rubber Duck Debugging: A Proposed Model

Building on the synergies identified, we propose an enhanced model of rubber duck debugging that incorporates AI assistance:

1.  Verbalization: The developer explains the code aloud, engaging metacognitive processes and stimulating subconscious thought patterns.

2.  AI Augmentation: As the developer verbalizes, the AI coding assistant (e.g., Copilot) processes these inputs along with the existing code context, generating real-time suggestions.

3.  Synergistic Insight Generation: The interplay between verbalization, subconscious processing, and AI suggestions leads to new insights. The AI's suggestions may trigger subconscious connections or prompt the developer to verbalize new aspects of the problem.

4.  Iterative Refinement: This process repeats iteratively, with each cycle potentially yielding deeper understanding or novel solutions. The developer's verbalizations become more refined and targeted, guided by subconscious insights and AI suggestions.

5.  Metacognitive Reflection: Periodically, the developer steps back to reflect on the process itself, assessing the evolution of their understanding and the contribution of the AI suggestions.

This model leverages the strengths of both human cognition and AI capabilities:

-              The human developer brings creativity, contextual understanding, and the ability to verbalize complex concepts.

-              The AI assistant contributes rapid pattern recognition, access to a vast knowledge base, and the ability to generate diverse code suggestions.

-              The rubber duck (or its conceptual equivalent) is a focal point for verbalization and externalized thinking.

5. Neurological and Cognitive Mechanisms

The proposed AI-augmented rubber duck debugging model engages multiple cognitive and neurological processes, leveraging various brain regions and networks to enhance problem-solving capabilities:

5.1 Language Processing and Production

The verbalization component of rubber duck debugging heavily involves Broca's and Wernicke's areas, critical regions for language processing and production (Flinker et al., 2015). This linguistic engagement may facilitate a more comprehensive processing of the code's logic and structure.

·       Broca's Area: Located in the frontal lobe, Broca's area is crucial for speech production. When developers verbalize their code, this region is highly active, potentially enhancing their ability to articulate complex logical structures.

·       Wernicke's Area: In the temporal lobe, Wernicke's area is responsible for language comprehension. As developers listen to their explanations and process AI suggestions, this area helps understand and integrate this information.

The interplay between these areas during verbalization may lead to:

·       Improved Code Comprehension: The act of translating code into natural language can highlight logical inconsistencies or overlooked details.

·       Enhanced Problem Reformulation: Verbalization may help developers reframe the problem in new ways, potentially leading to novel solutions.

5.2 Working Memory and Cognitive Load

Interaction with the AI assistant and verbalization may serve as forms of cognitive offloading, freeing up working memory resources for higher-level problem-solving (Risko & Gilbert, 2016). However, integrating AI suggestions also introduces new information, requiring careful cognitive load management.

Key aspects of this mechanism include:

·       Cognitive Offloading: Developers can reduce the strain on their working memory by externalizing thoughts through verbalization and relying on AI for specific mental tasks (e.g., recalling syntax or design patterns).

·       Dynamic Resource Allocation: As some cognitive resources are freed up, they can be reallocated to more complex problem-solving aspects, such as high-level design decisions or algorithmic optimization.

·       Balanced Information Integration: While AI suggestions provide valuable input, they also require processing. Developers must learn to balance the cognitive load of integrating AI suggestions with the benefits they provide.

For example, a developer working on a complex data structure might offload the task of remembering exact syntax to the AI, allowing them to focus more on the overall structure and efficiency of the algorithm.

5.3 Default Mode Network Activation

The creative problem-solving aspects of this process may engage the Default Mode Network (DMN), a brain network associated with self-referential thinking, creativity, and insight generation (Beaty et al., 2016). The DMN includes regions such as the medial prefrontal cortex, posterior cingulate cortex, and angular gyrus.

The interplay between focused verbalization and the generative suggestions of the AI may create optimal conditions for DMN activation and insight generation:

·       Incubation and Insight: Periods of verbalization followed by reflection on AI suggestions may provide the right balance of focus and relaxation that often precedes creative insights.

·       Associative Processing: The DMN's role in making novel connections between disparate ideas may be enhanced by the diverse suggestions provided by AI.

·       Self-Referential Thinking: As developers explain their code and consider AI suggestions, they may engage in metacognitive processes, reflecting on their problem-solving strategies and biases.

5.4 Predictive Processing

Interaction with the AI assistant may enhance the brain's predictive processing mechanisms (Clark, 2013). As developers become accustomed to the AI's suggestion patterns, their brains may develop more sophisticated predictive models, potentially leading to faster recognition of functional code patterns and solutions.

This predictive processing enhancement may manifest in several ways:

·       Improved Pattern Recognition: Regular exposure to AI-suggested patterns may train developers to recognize these patterns more quickly in future coding sessions, even without AI assistance.

·       Anticipatory Thinking: Developers may begin to anticipate the kinds of suggestions the AI might make, leading to more proactive problem-solving approaches.

·       Error Prediction: Enhanced predictive processing may improve developers' ability to foresee potential errors or edge cases in their code.

For instance, after multiple sessions of working with an AI assistant, a developer might start to internalize standard optimization techniques suggested by the AI, applying them preemptively in future coding tasks.

By engaging these various neurological and cognitive mechanisms, AI-augmented rubber duck debugging can significantly enhance developers' problem-solving capabilities, leading to more efficient and creative software development processes.

6. Implications and Future Directions

Integrating AI coding assistants with rubber duck debugging and other metacognitive techniques has far-reaching implications for software development and cognitive enhancement. This section explores these implications in detail and suggests future research and development directions.

6.1 Enhanced Developer Productivity and Creativity

The synergistic interaction between human cognition and AI assistance has the potential to enhance developer productivity and creativity in several ways significantly:

1.        Accelerated Code Generation: AI suggestions can speed up the process of writing boilerplate code, allowing developers to focus on more complex and creative aspects of programming.

2.        Reduced Debugging Time: AI-augmented rubber duck debugging could significantly reduce the time spent debugging by helping identify potential issues early in the development process.

3.        Expanded Solution Space: AI suggestions may introduce developers to novel approaches or techniques they hadn't considered, potentially leading to more innovative solutions.

4.        Optimized Workflow: Integrating AI into the development process may lead to new, more efficient workflows that combine the strengths of human insight and machine processing.

5.        Enhanced Code Quality: AI assistance in identifying best practices and potential improvements may increase the overall quality of code produced.

Quantitative studies measuring productivity gains and qualitative assessments of solution creativity will be crucial to understand this approach's impact fully.

6.2 Skill Development and Learning

Regular interaction with AI-augmented rubber duck debugging may accelerate the development of advanced programming skills:

1.        Pattern Internalization: Exposure to AI-suggested patterns and solutions may help developers quickly internalize complex coding patterns.

2.        Best Practice Absorption: Consistent interaction with AI suggestions aligned with industry best practices could accelerate developers' adoption of these practices.

3.        Cross-Language Skill Transfer: AI assistants' ability to work across multiple programming languages may facilitate skill transfer between languages for developers.

4.        Accelerated Novice-to-Expert Transition: Novice programmers can learn from AI suggestions, potentially accelerating their journey to becoming expert developers.

5.        Continuous Learning: AI models' ever-updating nature means developers can be continuously exposed to new techniques and practices, fostering a culture of lifelong learning.

Longitudinal studies tracking developer skill progression with and without AI assistance would provide valuable insights into these learning effects.

6.3 Metacognitive Enhancement

The process of AI-augmented rubber duck debugging may foster stronger metacognitive skills among developers:

1.        Enhanced Self-Reflection: The need to articulate thoughts to both the "rubber duck" and the AI assistant may encourage deeper self-reflection on one's problem-solving processes.

2.        Improved Thought Externalization: Regular practice in explaining code and problems may enhance developers' ability to externalize their thought processes effectively.

3.        Critical Evaluation Skills: The need to evaluate AI suggestions may sharpen developers' critical thinking skills, improving their ability to assess both their own ideas and external input.

4.        Increased Awareness of Knowledge Gaps: Interaction with AI assistants may help developers become more aware of areas where their knowledge is lacking, promoting targeted learning.

5.        Enhanced Problem Decomposition: The process of explaining problems for AI assistance may improve developers' ability to break down complex problems into manageable components.

Research into how these metacognitive skills develop and transfer to other domains could provide valuable insights into cognitive enhancement strategies.

6.4 Ethical Considerations

As AI assistants become more sophisticated, several ethical considerations come to the forefront:

1.        Authorship and Intellectual Property: Questions about code ownership and authorship may arise when significant portions of code are suggested by AI (Bryson & Winfield, 2017).

2.        Over-Reliance on AI: Developers may become overly dependent on AI suggestions, stunting their own skill development.

3.        Bias in AI Suggestions: AI models may perpetuate biases in their training data, potentially propagating problematic coding practices or solutions.

4.        Privacy Concerns: AI assistants may raise questions about the privacy of the code being analyzed and the data used to train these models.

5.        Impact on Employment: The increasing capabilities of AI coding assistants may impact the job market in software development.

As AI technologies become more prevalent, developing ethical guidelines and best practices for using AI in software development will be crucial.

6.5 Generalization to Other Domains

The principles of AI-augmented metacognition explored in this study may have applications beyond software development:

1.        Scientific Research: Similar approaches could be applied to hypothesis generation and experimental design in scientific research.

2.        Creative Writing: AI assistants could aid in the brainstorming and drafting processes in creative writing.

3.        Engineering Design: Complex engineering problems might benefit from AI-augmented problem-solving techniques.

4.        Business Strategy: Strategic planning and problem-solving in business contexts could be enhanced by similar AI-augmented metacognitive approaches.

5.        Education: The principles of AI-augmented learning could be applied to develop new educational tools and techniques across various disciplines.

Exploring how these principles can be adapted and applied in other fields could lead to significant advancements in problem-solving and creativity across multiple domains.

7. Future Research Directions

Several avenues for future research emerge from this study:

1.        Empirical validation of the proposed model through controlled studies comparing traditional rubber duck debugging with the AI-augmented approach. This could involve measuring metrics such as time to solution, code quality, and developer satisfaction across different problem types and difficulty levels.

2.        Neuroimaging studies to map brain activity during AI-augmented rubber duck debugging, potentially revealing new insights into the neurological basis of human-AI collaborative cognition. This could use techniques such as fMRI or EEG to observe changes in brain activation patterns during different stages of the debugging process.

3.        Longitudinal studies on the impact of prolonged use of AI-augmented debugging techniques on developer skills, intuition, and problem-solving abilities. This could track developers over months or years, assessing how their coding practices and cognitive approaches evolve with consistent AI assistance.

4.        Development of specialized AI tools explicitly designed to enhance metacognitive techniques in programming and other domains. This might involve creating AI models that are trained not just on code but also on problem-solving and verbalization in programming contexts.

5.        Exploration of potential risks, such as cognitive biases or over-reliance on AI suggestions, and development of mitigation strategies. This could include studying how developers' decision-making processes change with AI assistance and developing guidelines or training programs to maintain critical thinking skills.

6.        Investigation into the optimal balance of AI assistance and independent problem-solving for skill development. This could involve experimental studies with varying levels of AI involvement to determine the most effective learning and skill retention approach.

7.        Cross-cultural studies on the effectiveness of AI-augmented rubber duck debugging, exploring how different verbal and cognitive styles across cultures interact with this technique.

8. Conclusion

Integrating AI coding assistants with traditional metacognitive techniques like rubber duck debugging represents a significant evolution in cognitive augmentation for programming tasks. Developers may achieve new productivity levels, insight, and creativity by leveraging the synergy between subconscious mental processes, verbalization techniques, and AI capabilities.

This AI-augmented approach to rubber duck debugging enhances existing practices and opens up new possibilities for human-AI collaboration in complex cognitive tasks. As we continue to explore and refine these techniques, we may be witnessing the emergence of a new paradigm in software development—one that optimally combines human creativity and intuition with the vast knowledge and pattern recognition capabilities of AI systems.

The journey of "quacking the code" with AI-augmented rubber duck debugging is just beginning, promising exciting advancements in cognitive enhancement, software development practices, and our understanding of human-AI collaborative cognition.

References

Beaty, R. E., Benedek, M., Silvia, P. J., & Schacter, D. L. (2016). Creative cognition and brain network dynamics. Trends in cognitive sciences, 20(2), 87-95.

Boden, M. A. (2004). The creative mind: Myths and mechanisms. Psychology Press.

Bryson, J. J., & Winfield, A. (2017). Standardizing ethical design for artificial intelligence and autonomous systems. Computer, 50(5), 116-119.

Chen, M., Tworek, J., Jun, H., Yuan, Q., Pinto, H. P. D. O., Kaplan, J., ... & Zaremba, W. (2021). Evaluating large language models trained on code. arXiv preprint arXiv:2107.03374.

Clark, A. (2013). Whatever next? Predictive brains, situated agents, and the future of cognitive science. Behavioral and brain sciences, 36(3), 181-204.

Flavell, J. H. (1979). Metacognition and cognitive monitoring: A new area of cognitive–developmental inquiry. American psychologist, 34(10), 906.

Flinker, A., Korzeniewska, A., Shestyuk, A. Y., Franaszczuk, P. J., Dronkers, N. F., Knight, R. T., & Crone, N. E. (2015). Redefining the role of Broca's area in speech. Proceedings of the National Academy of Sciences, 112(9), 2871-2875.

Hunt, A., & Thomas, D. (1999). The pragmatic programmer: from journeyman to master. Addison-Wesley Professional.

Kihlstrom, J. F. (1987). The cognitive unconscious. Science, 237(4821), 1445-1452.

Kounios, J., & Beeman, M. (2009). The Aha! moment: The cognitive neuroscience of insight. Current directions in psychological science, 18(4), 210-216.

Miyake, N., & Norman, D. A. (1979). To ask a question, one must know enough to know what is not known. Journal of verbal learning and verbal behavior, 18(3), 357-364.

Risko, E. F., & Gilbert, S. J. (2016). Cognitive offloading. Trends in cognitive sciences, 20(9), 676-688.

Veissière, S. (2016). Varieties of tulpa experiences: The hypnotic nature of human sociality, personhood, and interphenomenality. Hypnosis and meditation: Towards an integrative science of conscious planes, 55-76.