Specification as Thinking
Slide Idea
The slide argues that explicit specification matters because it increases traceability and accountability in creative and technical work. Pre-production documents such as scripts, shot lists, and prompts function as cognitive artifacts that externalize decisions, making previously internal thinking visible and reviewable. The slide emphasizes that AI systems do not make autonomous decisions but rather require human decision-makers to articulate their intentions explicitly before system execution can occur.
Key Concepts & Definitions
Specification
Specification is the explicit articulation of requirements, intentions, or design decisions in documented, communicable form. In creative and technical contexts, specifications translate tacit understanding or internal vision into concrete descriptions that others—collaborators, tools, systems—can interpret and act upon. Specifications range from high-level objectives ("create tension in this scene") to detailed technical parameters ("use 50mm lens at f/2.8, position camera 3 feet from subject"). The act of specifying forces decision-makers to clarify ambiguous intentions, resolve internal contradictions, and make implicit assumptions explicit.
Source: Cross, N. (2011). Design Thinking: Understanding How Designers Think and Work. Berg Publishers.
Traceability
Traceability refers to the documented ability to follow the relationships between decisions, requirements, implementations, and outcomes throughout a project lifecycle. In production contexts, traceability means being able to trace backward from a final outcome to the specifications and decisions that produced it, or forward from initial requirements to verify that they were implemented. Traceability requires explicit documentation rather than relying on memory or tacit understanding. When specifications exist, practitioners can answer questions like "Why did we make this choice?" or "Which requirement does this feature address?" with reference to documented artifacts rather than post-hoc rationalization.
Source: Raji, I. D., Smart, A., White, R. N., Mitchell, M., Gebru, T., Hutchinson, B., Smith-Loud, J., Theron, D., & Barnes, P. (2020). "Closing the AI Accountability Gap: Defining an End-to-End Framework for Internal Algorithmic Auditing." In Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency (pp. 33-44). ACM.
Accountability
Accountability is the condition of being answerable for decisions, actions, and outcomes, with the ability to explain and justify the reasoning behind choices made. In creative and technical practice, accountability depends on documentation that records who made which decisions, when, based on what criteria or constraints. Without explicit specifications, accountability becomes difficult because no record exists of the decision-making process—only the final output. Accountability is particularly critical in collaborative work where multiple practitioners contribute to outcomes, in professional contexts where decisions must be defended, and in systems where outcomes affect stakeholders who were not present during decision-making.
Source: Raji, I. D., et al. (2020). "Closing the AI Accountability Gap: Defining an End-to-End Framework for Internal Algorithmic Auditing." Conference on Fairness, Accountability, and Transparency.
Cognitive Artifacts
Cognitive artifacts are external objects or systems designed to maintain, display, or operate upon information in ways that support representational functions and affect human cognitive performance. Examples include written lists, diagrams, maps, scripts, spreadsheets, and digital tools. Cognitive artifacts do not simply store information passively—they transform the nature of cognitive tasks by externalizing mental work into manipulable, shareable, persistent forms. A shot list, for instance, transforms the mental task of remembering planned coverage into the physical task of consulting documented specifications. This externalization reduces cognitive load, enables collaborative review, creates permanent records, and allows iterative refinement of thinking.
Source: Norman, D. A. (1991). "Cognitive Artifacts." In J. M. Carroll (Ed.), Designing Interaction: Psychology at the Human-Computer Interface (pp. 17-38). Cambridge University Press.
Pre-Production Documentation
Pre-production documentation encompasses the collection of written, drawn, or notated specifications created before production execution begins. In filmmaking, this includes scripts (dialogue and action descriptions), shot lists (planned camera setups), storyboards (visual sequence planning), location scouts (spatial documentation), schedules (temporal planning), and budget documents (resource allocation). These documents serve multiple functions simultaneously: they externalize creative vision, coordinate collaborator understanding, establish shared reference points, create accountability records, and enable revision before resource-intensive production occurs. Pre-production documents are not merely planning aids but cognitive artifacts that make thinking visible and improvable.
Source: Bordwell, D., & Thompson, K. (2017). Film Art: An Introduction (11th ed.). McGraw-Hill Education.
System Autonomy vs. Human Decision-Making
In computational contexts, the distinction between system autonomy and human decision-making centers on where intentionality and judgment reside. AI systems, despite sophisticated pattern recognition and output generation capabilities, do not possess intentions or make value-based judgments autonomously. They execute operations based on prior specifications—training data, model architectures, hyperparameters, prompts, constraints. Human practitioners make decisions about what objectives to pursue, which trade-offs to accept, which outcomes qualify as acceptable. The system requires these human decisions to be articulated explicitly as inputs, parameters, or prompts before it can function. The appearance of system "decision-making" actually reflects prior human specification work made invisible by automation.
Source: European Commission. (2025). "Guidelines on the Definition of an AI System Under the AI Act."
Why This Matters for Students' Work
Understanding specification as thinking rather than as mere documentation fundamentally changes how students approach creative and technical work. Many students treat specification documents—scripts, shot lists, design briefs, technical requirements—as administrative overhead imposed by instructors or professional conventions. Recognizing that specification is itself a cognitive process that clarifies, tests, and improves thinking transforms these documents from bureaucratic requirements into valuable thinking tools.
The traceability benefit becomes particularly important in revision and evaluation contexts. When students create explicit specifications before execution, they establish criteria against which to assess outcomes. A filmmaker with a documented shot list can diagnose why coverage feels incomplete—specific planned shots were never captured or were executed differently than specified. A designer with written specifications can evaluate whether implemented features actually address stated requirements. Without specification documents, students can only evaluate outcomes based on vague impressions rather than explicit criteria. The specification creates the standard against which work can be measured.
Accountability extends beyond avoiding blame to enabling learning from outcomes. When students document their decision-making rationale—"I chose this lighting approach to create mood X" or "I structured the argument this way to prioritize accessibility over technical precision"—they create records they can review after completion. These records reveal patterns in their decision-making: which kinds of specifications they habitually overlook, which assumptions prove incorrect, which trade-offs they consistently regret. Without documented specifications, students cannot reliably learn from experience because they cannot accurately remember or reconstruct what they actually decided versus what they retrospectively wish they had decided.
In collaborative production, explicit specification becomes essential for coordination. When multiple students work on shared projects, tacit understanding does not scale—each person's internal model of "what we're trying to achieve" diverges over time without explicit reference points. Specifications create shared understanding that persists independent of any individual's memory. They enable asynchronous collaboration where contributors work at different times, distributed collaboration where team members have different expertise levels, and handoff scenarios where work passes between people who may never directly communicate.
The cognitive artifact dimension reveals specification as thinking support rather than mere recording. The act of writing a specification often reveals problems in initial thinking: contradictions between stated objectives, missing details necessary for execution, assumptions that seem obvious internally but are actually questionable when externalized. Students frequently discover that what seemed clear in their minds becomes incoherent when articulated explicitly. This discovery is valuable—it means specification catches conceptual problems before they become execution failures. The document serves as a thinking aid, not just a communication tool.
For work involving computational systems—AI tools, generative models, automated processes—the requirement for explicit specification becomes non-negotiable. These systems cannot access unstated intentions. A prompt that says "make it interesting" fails not because the system is inadequate but because "interesting" has not been specified in executable terms. Students must learn to articulate aesthetic judgments, functional requirements, and success criteria with precision sufficient that systems lacking human interpretive capabilities can operationalize them. This requirement forces unprecedented clarity in specification work.
The accountability dimension becomes particularly significant in professional and ethical contexts. When work affects others—users, clients, audiences, communities—being able to explain and justify decisions requires documented specifications. Students cannot defend design choices, argue for resource allocations, or explain why specific approaches were selected without records of the reasoning that guided those choices. Professional practice increasingly demands documentation not just of what was decided but why, based on which criteria, considering which alternatives. Specification documents provide the evidence base for accountability.
How This Shows Up in Practice (Non-Tool-Specific)
Filmmaking Pre-Production
A director planning a scene may have a clear internal vision of the intended emotional effect but must translate that vision into concrete specifications for the production team. This requires creating shot lists specifying camera positions, lens choices, movements, and framing for each planned setup. Lighting diagrams specify instrument positions, intensities, and color temperatures. These specifications serve multiple functions: they coordinate understanding across cinematographer, gaffer, and camera operator; they create a record enabling the team to verify that planned coverage was actually captured; they document decision rationale when reviewing footage later. The specifications make the director's tacit vision explicit and executable.
Design Requirements Documentation
A designer working on an interface must convert user needs and business objectives into explicit functional and aesthetic specifications. This documentation might include user flows (specifying navigation paths), wireframes (specifying layout and hierarchy), interaction specifications (specifying system responses to user actions), and accessibility requirements (specifying support for assistive technologies). These specifications enable developers to implement designs accurately, provide criteria for quality assurance testing, create accountability records showing how design decisions address stated requirements, and support future designers who may need to modify or extend the system without access to the original designer's tacit understanding.
Writing and Editorial Standards
A writer collaborating with editors must often make style and structural decisions explicit through specification documents. This includes style guides specifying voice, tone, terminology, and formatting conventions; outlines specifying argument structure and evidence sequencing; revision notes documenting which changes were made and why. In academic contexts, citation specifications document evidentiary basis for claims. These specifications enable multiple writers to maintain consistency across collaborative documents, provide editors with criteria for evaluating whether drafts meet objectives, create accountability for factual accuracy and source attribution, and support revision by making previous decision rationale available for reconsideration.
Engineering Design Documentation
Engineers developing systems must create specifications at multiple levels: requirements documents specifying what the system must accomplish, architecture documents specifying how major components interact, interface specifications specifying how subsystems communicate, and test specifications defining success criteria. These documents enable team members with different expertise to work on integrated components, provide traceability showing which design elements address which requirements, create accountability records for safety-critical decisions, and support maintenance engineers who must understand system behavior without access to original designers. Specification documents persist beyond any individual's tenure on the project.
Computational System Prompts and Parameters
When using generative AI systems, practitioners must specify desired outcomes through prompts, parameters, examples, or constraints. A prompt requesting "generate marketing copy" produces generic results because objectives, audience, tone, length, key messages, and constraints remain unspecified. Effective specification requires articulating: target audience characteristics, desired emotional response, key information to include, length constraints, stylistic preferences, brand voice parameters. This specification work externalizes decision-making that remains invisible in direct human creation but becomes an explicit requirement when working through systems that cannot infer unstated intentions.
Collaborative Production Handoffs
In production workflows where work passes between specialists—writer to editor, designer to developer, cinematographer to colorist—specifications enable coordination without requiring continuous synchronous communication. A colorist working from a cinematographer's notes can understand intended mood and technical constraints. A developer implementing from design specifications can make appropriate trade-off decisions when perfect implementation proves infeasible. These specifications create persistent, shareable records that remain accessible even when original decision-makers are unavailable. They transform tacit knowledge held by individuals into explicit knowledge accessible to teams.
Common Misunderstandings
"Specifications are just administrative requirements that don't contribute to creative quality"
This fundamentally misunderstands specification as cognitive activity. The act of articulating specifications forces clarification of ambiguous intentions, reveals contradictions in initial thinking, and identifies missing considerations before resource-intensive execution begins. High-quality specifications correlate with high-quality outcomes not because documentation itself improves execution but because the thinking required to produce clear specifications improves decision-making. Students who view specification as bureaucratic overhead miss the cognitive benefits—they produce vague documents that fulfill formal requirements without actually thinking through their projects systematically.
"Good practitioners work intuitively without needing to document their thinking"
This confuses individual expertise with collaborative practice requirements. Even highly experienced practitioners who can execute effectively based on tacit judgment still create specifications in professional contexts for coordination, accountability, and knowledge transfer purposes. The difference is that experts can produce specifications more efficiently and recognize which aspects require detailed articulation versus which can remain implicit. But the notion that expert practice eliminates specification needs is incorrect—professional production environments demand explicit documentation precisely because tacit understanding does not scale across teams, time, or organizational changes.
"AI systems make autonomous decisions based on their training"
This misattributes agency and intentionality to computational processes. AI systems execute operations based on specifications embedded in their design: training data selection, model architecture choices, hyperparameter settings, prompt inputs, constraint definitions. Every output reflects prior human decisions about objectives, evaluation criteria, and acceptable trade-offs. The system does not "decide" which kind of text to generate—it generates text that maximizes probability distributions learned from training data selected by humans to approximate human-specified objectives. Attributing decision-making to systems obscures the extensive human specification work that precedes and enables system operation.
"Specifications limit creativity by over-determining outcomes before exploration"
This reverses the actual relationship between specification and creative exploration. Specifications define objectives, constraints, and success criteria—not the specific solutions that meet those criteria. A shot list specifies what coverage to capture, not the only possible way to capture it. Requirements documents specify what the system must accomplish, not the only design that could accomplish it. Specifications bound the solution space (which constraints must be respected) while leaving methods for meeting requirements open to creative problem-solving. Students who avoid specification often produce unfocused work that lacks clear objectives, not work that exhibits more creative freedom.
Scholarly Foundations
Cross, N. (2011). Design Thinking: Understanding How Designers Think and Work. Berg Publishers.
Empirical analysis of expert design practice across multiple domains showing how designers externalize thinking through sketches, diagrams, notations, and specifications. Demonstrates that specification is not documentation of completed thinking but active cognitive process that develops and refines understanding.
Raji, I. D., Smart, A., White, R. N., Mitchell, M., Gebru, T., Hutchinson, B., Smith-Loud, J., Theron, D., & Barnes, P. (2020). "Closing the AI Accountability Gap: Defining an End-to-End Framework for Internal Algorithmic Auditing." In Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency (pp. 33-44). ACM.
Establishes framework for accountability in AI system development through documentation of decision-making throughout development lifecycle. Argues that traceability—ability to trace outcomes back to specifications and decisions—is prerequisite for accountability. Demonstrates that without explicit specifications, accountability becomes impossible.
Schön, D. A. (1983). The Reflective Practitioner: How Professionals Think in Action. Basic Books.
Foundational analysis of professional practice showing how practitioners externalize thinking through "design moves"—specifications and implementations that test understanding and reveal new insights. Introduces concept of reflection-in-action and reflection-on-action, both enabled by externalizing decisions into observable form. Relevant to understanding specification as cognitive artifact supporting reflective practice.
Norman, D. A. (1991). "Cognitive Artifacts." In J. M. Carroll (Ed.), Designing Interaction: Psychology at the Human-Computer Interface (pp. 17-38). Cambridge University Press.
Foundational work establishing cognitive artifacts as external representations that transform the nature of cognitive tasks rather than merely supporting them. Analyzes how written specifications, diagrams, and notational systems change thinking by externalizing mental operations into manipulable external forms. Critical theoretical foundation for understanding specification documents as cognitive tools.
Bordwell, D., & Thompson, K. (2017). Film Art: An Introduction (11th ed.). McGraw-Hill Education.
Comprehensive analysis of filmmaking practice emphasizing role of pre-production planning and specification in coordinating complex collaborative production. Examines scripts, storyboards, shot lists as essential tools for translating creative vision into executable production plans. Demonstrates how specification documents enable coordination across specialized roles.
Kirsh, D. (1995). "The Intelligent Use of Space." Artificial Intelligence, 73(1-2), 31-68.
Analyzes how external spatial arrangements and notations serve cognitive functions by reducing internal computation load, providing persistent memory, and enabling pattern recognition through physical manipulation. Demonstrates that externalization fundamentally changes cognitive processes. Relevant to understanding how specification documents function as cognitive scaffolding.
Hutchins, E. (1995). Cognition in the Wild. MIT Press.
Ethnographic analysis of navigation practice showing how cognitive work is distributed across practitioners and artifacts. Demonstrates that complex cognitive tasks are accomplished through coordination of humans with representational artifacts like charts, logs, and instruments. Establishes distributed cognition framework relevant to understanding specification as distributed thinking support.
Zhang, J., & Norman, D. A. (1994). "Representations in Distributed Cognitive Tasks." Cognitive Science, 18(1), 87-122.
Theoretical analysis distinguishing internal versus external representations and demonstrating how external representations transform task structure. Shows that tasks performed with external representations differ fundamentally from same tasks performed mentally. Provides cognitive science foundation for understanding specification as thinking transformation.
Suchman, L. A. (2007). Human-Machine Reconfigurations: Plans and Situated Actions (2nd ed.). Cambridge University Press.
Critical analysis of relationship between plans/specifications and actual practice. Argues that specifications do not deterministically control action but serve as resources for situated action. Demonstrates that specifications must be interpreted and adapted during execution. Important for understanding limits and appropriate uses of specification documents.
Boundaries of the Claim
This slide does not claim that all thinking can or should be completely externalized into specifications. Significant aspects of expert practice remain tacit, embodied, and resistant to full articulation. The claim is that explicit specification increases traceability and accountability for those aspects of decision-making that can be articulated, not that specification eliminates the role of tacit judgment or improvisation.
The slide does not claim that specifications eliminate the need for interpretation during execution. Even detailed specifications require situated judgment about how to apply them in specific circumstances, particularly when unforeseen complications arise. Specifications guide and constrain action but do not mechanistically determine it.
This slide does not claim that more detailed specification always produces better outcomes. The appropriate level of specification granularity depends on project complexity, collaboration structure, risk tolerance, and resource constraints. Over-specification can create rigidity that prevents adaptive response to emergent insights during production.
The statement about AI systems not making decisions does not claim that all computational processes are equivalent or that AI systems lack sophistication. It clarifies that intentionality and value-based judgment reside with human practitioners who specify objectives and constraints, not with systems that optimize for those specifications. This distinction matters for understanding where responsibility and accountability appropriately reside.
This slide intentionally leaves open questions about which aspects of creative practice most benefit from explicit specification versus remaining tacit, about how specification practices should evolve as computational tools become more sophisticated, and about cultural and contextual variations in specification norms across different creative and technical domains.
Reflection / Reasoning Check
Reflection Question 1:
Consider a recent project where difficulty was experienced during execution or disappointment felt with final outcomes. To what extent were explicit specifications created before beginning work—documenting objectives, success criteria, planned approaches, or decision rationale? If more detailed specifications had been created, which execution difficulties might have been prevented or diagnosed more easily? Which aspects of initial thinking would specification have forced into clarity before problems emerged?
Reflection Question 2:
Think about a situation where a creative or technical decision needed explanation or justification—perhaps in peer review, client presentation, or evaluation context. How easy was it to reconstruct decision-making rationale after the fact? Would documented specifications created during the decision-making process have made accountability easier or more credible? What kinds of specifications would have provided useful evidence for explaining why particular choices were made under specific constraints?