15. Graphics Engineering, Governance, and Future Directions

Introduction and Context

Computer graphics is often presented as a technical discipline concerned with modelling, rendering, animation, visualization, and interaction. In professional practice, however, a successful graphics system is not defined only by image quality or rendering speed. It must also be planned, financed, maintained, tested, governed, legally defensible, ethically responsible, and adaptable to changing platforms. A visually impressive product can still fail if its assets are untraceable, its build pipeline is unreliable, its licensing assumptions are incorrect, or its synthetic content misleads users.

Graphics engineering therefore extends beyond algorithms and APIs. It combines software engineering, production management, design governance, platform strategy, information security, intellectual-property management, and responsible innovation. This perspective is important in game development, visual effects, computer-aided design, scientific visualization, virtual production, simulation, augmented reality, digital twins, and generative media.

The strategic question is not simply “How can an image be rendered?” It is “What graphics capability should an organization develop, acquire, govern, and sustain?” This chapter examines that question through project planning, production pipelines, build-versus-buy decisions, standards, quality assurance, governance, ethics, sustainability, and emerging research.

Learning Objectives

After studying this chapter, students should be able to:

1. Plan a graphics project using appropriate scope, milestones, resources, risks, and quality criteria.

2. Explain the structure and governance of a modern graphics production pipeline.

3. Evaluate build-versus-buy and platform decisions using technical, financial, organizational, and strategic criteria.

4. Design asset, metadata, version-control, testing, and standards practices for graphics projects.

5. Identify intellectual-property, privacy, bias, authenticity, safety, and regulatory concerns in visual computing.

6. Assess the environmental and operational sustainability of graphics systems.

7. Analyze emerging developments such as neural rendering, differentiable graphics, spatial computing, and synthetic media.

Graphics Engineering as a Systems Discipline

A graphics project produces more than pictures. It produces a coordinated system of assets, software, data, workflows, interfaces, documentation, and decisions. A useful engineering model divides the system into five interacting layers.

The creative layer contains visual concepts, narratives, layouts, materials, motion, lighting, and user experience. The representation layer contains meshes, textures, shaders, animations, scene graphs, point clouds, volumetric data, and metadata. The computational layer includes rendering engines, GPU programs, simulation systems, image-processing operations, and machine-learning models. The delivery layer addresses target devices, operating systems, network conditions, installation, deployment, and runtime monitoring. The governance layer establishes ownership, standards, approval authority, security, compliance, and accountability.

Failure in one layer can compromise the others. A highly detailed model may be unusable on a mobile device. A fast renderer may produce inconsistent results because color management is uncontrolled. A generative model may create visually convincing assets whose training provenance is unknown. Graphics engineering connects these concerns before they become expensive defects.

Several terms are central to this discipline. An asset is a reusable visual, geometric, auditory, or descriptive element. A pipeline is the sequence of activities and tools through which assets and code move from conception to delivery. A build is a reproducible assembly of source code, assets, configuration, and dependencies. Provenance records the origin and transformation history of an asset or dataset. Interoperability is the ability of different tools or systems to exchange and correctly interpret information. Technical debt is the future cost created by expedient but inadequate technical decisions.

From Creative Brief to Executable Plan

Planning begins with a creative and technical brief. The brief should describe the intended audience, visual objectives, interaction model, platforms, performance targets, accessibility requirements, delivery dates, constraints, and acceptance criteria. Vague goals such as “cinematic quality” should be translated into measurable or reviewable conditions: target frame rate, resolution, latency, color range, polygon or texture budgets, supported devices, and approval milestones.

A graphics project is often best managed as a network of interdependent work packages rather than as a single linear task. Concept development may precede modelling, but shader development can proceed in parallel with asset creation. Platform testing should begin before final production because device limitations may alter modelling, lighting, or interface decisions. A work-breakdown structure can identify activities such as concept design, asset acquisition, modelling, UV layout, texturing, rigging, animation, simulation, lighting, compositing, integration, optimization, testing, and release.

Milestones should represent decisions, not merely elapsed time. Examples include approval of the visual language, completion of a representative vertical slice, technical validation on target hardware, content lock, and release candidate approval. A vertical slice is a small but integrated sample that demonstrates the intended quality and pipeline across several stages. It is more informative than producing many isolated assets because it reveals interactions among tools, performance, visual style, and review processes.

Resource planning must include specialists, software, hardware, storage, rendering capacity, external vendors, and contingency. Schedule estimates should account for iteration and review. Graphics work is particularly vulnerable to late changes because an alteration to topology, scale, coordinate systems, or color management can propagate through many downstream assets. Change-control procedures should therefore distinguish creative changes, technical corrections, compliance changes, and emergency fixes.

Risk registers provide a structured way to identify threats. Typical risks include unavailable hardware, unstable third-party plugins, unclear asset rights, data loss, performance failure, model bias, security vulnerabilities, and dependence on a single vendor. Each risk should have an owner, likelihood, impact, mitigation, trigger, and contingency response.

The Production Pipeline and Its Controls

A professional pipeline commonly begins with pre-production, where requirements, visual references, conventions, prototypes, and technical constraints are established. Production then creates or acquires assets and code. Integration assembles those elements into scenes, applications, or experiences. Verification evaluates correctness, performance, usability, security, and compliance. Release packages the system for its intended platforms, while operations monitor it after deployment.

The pipeline should be treated as a controlled flow of information. Each transition needs an agreed input, output, responsible role, and validation rule. For example, an approved model may need a specified coordinate system, naming convention, scale, polygon budget, material assignment, collision representation, and metadata package before it enters the integration stage.

Version control is essential for both code and binary assets. Text-based source files are well suited to conventional branching and merging, whereas large binary files require specialized repositories, locking, or content-addressed storage. Every release should be reproducible from identified versions of code, assets, tools, plugins, shaders, and configuration. Continuous integration can automatically compile shaders, validate asset schemas, run unit tests, render reference scenes, and compare output images against approved baselines.

Automation improves consistency but does not remove the need for judgment. A validation script can detect missing textures, invalid normals, excessive draw calls, or incorrect naming. It cannot always determine whether a composition communicates the intended meaning or whether a representation is culturally inappropriate. Effective pipelines combine automated gates with expert review.

Color management deserves special attention. Different displays, file formats, transfer functions, and working spaces can produce inconsistent results. A governed pipeline defines color spaces, transforms, reference monitors, export settings, and review conditions. Similar discipline is needed for units, coordinate handedness, frame rates, time codes, physical scale, texture compression, and level-of-detail rules.

Build, Buy, License, or Partner?

The build-versus-buy decision is not a simple comparison of purchase price and development cost. Organizations may build a rendering feature, purchase an engine, license an asset library, use a cloud service, or partner with a specialist studio. The choice should be assessed across several dimensions: strategic differentiation, total cost of ownership, schedule, performance, customization, interoperability, security, vendor dependence, support, legal rights, and internal capability.

Building is attractive when the capability is central to competitive advantage, when existing products cannot meet critical requirements, or when long-term control is essential. It is costly because the organization must maintain documentation, testing, compatibility, security, recruitment, and future upgrades. Buying or licensing can accelerate delivery and provide mature tooling, but it may impose subscription costs, usage restrictions, data-transfer charges, limited customization, or dependence on a vendor’s roadmap.

A useful decision process is to define requirements, identify alternatives, estimate total lifecycle cost, evaluate risks, conduct a proof of concept, and document the decision. Lifecycle cost includes acquisition, integration, training, support, infrastructure, migration, renewal, compliance, and eventual replacement. A technically superior option may be strategically poor if it prevents data export or makes the organization dependent on an inaccessible proprietary format.

Platform selection requires similar analysis. A native desktop application may provide high performance and extensive GPU access but reduce portability. A web-based system improves reach and deployment simplicity but may face browser, bandwidth, and security constraints. Mobile, console, immersive, cloud-rendered, and edge platforms each impose different requirements for input, thermal limits, memory, latency, privacy, and certification.

Architecture should separate stable domain concepts from platform-specific implementation. Abstraction layers, portable asset formats, documented interfaces, and automated conversion reduce migration costs. Portability should not be assumed merely because two systems support the same nominal file format; semantics, color, materials, animation behavior, and metadata may still differ.

Standards, Assets, and Quality Assurance

Standards are organizational agreements that reduce ambiguity. They may govern naming, folder structures, metadata, file formats, material definitions, scene scales, camera conventions, review states, and delivery packages. Standards should be specific enough to support automation but flexible enough to accommodate legitimate creative variation.

Asset governance begins with an asset register. Each entry may include a unique identifier, creator, source, version, dependencies, license, permitted uses, sensitivity classification, creation date, review status, and archival location. Provenance is especially important when assets are generated, purchased, scanned, captured from people, or derived from public datasets.

Quality assurance should operate at several levels. Unit testing checks isolated functions such as geometry conversion or color transformation. Integration testing examines interactions among tools, assets, and subsystems. Regression testing determines whether a change has damaged previously accepted behavior. Performance testing measures frame time, memory, loading, thermal behavior, network usage, and scalability. Visual testing compares rendered output with references while allowing controlled tolerances for floating-point or hardware variation. User testing evaluates comprehension, comfort, accessibility, and task success.

Graphics systems also need failure-oriented testing. Malformed files, missing dependencies, extreme model complexity, unusual lighting, incorrect permissions, corrupted caches, and interrupted downloads should be tested deliberately. For immersive systems, motion sickness, tracking loss, boundary violations, and user fatigue are relevant quality concerns. For synthetic-media systems, tests should include identity misuse, misleading edits, demographic performance differences, and failure to communicate uncertainty.

Governance, Intellectual Property, and Privacy

Governance answers who may create, approve, modify, release, archive, or delete graphics content. A practical governance model assigns responsibility to roles such as product owner, technical director, art director, pipeline engineer, security lead, legal adviser, and data-protection officer. Decision rights should be visible rather than inferred from organizational hierarchy.

Intellectual-property analysis must distinguish ownership, licensing, attribution, moral rights, publicity rights, trademark concerns, and contractual restrictions. A purchased model may permit use in one application but prohibit resale as a standalone asset. A photograph may involve rights held by the photographer, depicted individuals, location owners, or trademark holders. Training or fine-tuning a generative system introduces additional questions about dataset authorization, output similarity, record keeping, and contractual terms.

Privacy is not limited to text databases. Faces, bodies, voices, movement patterns, spatial scans, eye-tracking data, and home environments can identify or profile people. Data minimization, purpose limitation, informed consent, access controls, retention limits, encryption, and deletion procedures should be designed into graphics systems. In spatial computing, a map of a user’s room can reveal sensitive information even when no conventional personal record is stored.

Governance also requires an audit trail. It should be possible to determine which data, model, asset, operator, and software version contributed to a released result. The degree of traceability should be proportional to risk. A laboratory visualization and a public political deepfake do not require identical controls, but both require accountable processes.

Bias, Authenticity, and Responsible Synthetic Media

Visual systems can reproduce or amplify social bias. Bias may enter through training data, annotation, sampling, design assumptions, evaluation metrics, or unequal access to technology. A face-generation system may perform differently across demographic groups; a virtual human may encode stereotypes; a procedural city model may represent some communities as more dangerous or less valuable. Technical accuracy alone does not establish fairness.

Responsible practice requires identifying affected groups, documenting intended and excluded uses, evaluating performance across relevant populations, and providing channels for correction or appeal. Evaluation should include both quantitative metrics and qualitative review by people with contextual knowledge. It is also important to ask whether the system should be used at all in a given decision context.

Authenticity concerns whether users can understand how a visual artifact was produced and whether it represents a real event, person, or place. Synthetic media is not inherently harmful; it supports dubbing, accessibility, education, simulation, entertainment, and creative exploration. Risk increases when synthetic content impersonates real people, conceals material alteration, manipulates public understanding, or removes meaningful consent.

Possible safeguards include provenance records, cryptographic signatures, visible or machine-readable disclosure, controlled access, watermarking where appropriate, identity verification, human review, and rapid correction procedures. No single safeguard is sufficient. Watermarks can be removed, provenance can be lost during editing, and detection tools can produce false positives. Responsible deployment combines technical measures, policy, user education, and institutional accountability.

Sustainability and Operational Responsibility

Graphics systems consume energy through content creation, rendering, training, storage, transmission, and display. Sustainability analysis should consider the full lifecycle rather than only the runtime application. High-resolution assets may increase storage and network costs; excessive shader complexity may increase device energy use; large machine-learning models may require substantial training and inference resources.

Practical measures include adaptive quality, level-of-detail management, efficient texture formats, cache reuse, render scheduling, hardware-appropriate algorithms, asset deduplication, and disposal or reuse of equipment. Cloud rendering is not automatically sustainable: its impact depends on utilization, data movement, energy sources, cooling, and workload scheduling. Teams should measure energy or resource consumption where feasible and balance visual fidelity against user value.

Sustainability also includes social and organizational dimensions. Fragile pipelines create avoidable overtime, inaccessible interfaces exclude users, and opaque automation can shift risk onto less powerful workers or communities. A responsible graphics organization treats maintainability, accessibility, worker well-being, and long-term knowledge retention as engineering concerns.

Applied Case: A Virtual Heritage Experience

Background

The following is an illustrative case. A university museum and a small technology company plan an interactive virtual reconstruction of a historic building. The experience will run in a browser and on an immersive headset. The team intends to combine laser scans, archival drawings, procedural reconstruction, and generated environmental imagery.

Situation and Problem

Early prototypes look convincing, but the project faces four problems. The scan data contains visitors whose faces and body movements were captured without a clear reuse plan. Several archival images have uncertain copyright status. The browser version exceeds memory limits, while the headset version causes discomfort during rapid camera movement. Finally, generated imagery introduces architectural details that cannot be supported by historical evidence.

Analysis

The team creates an asset register and classifies every source as verified, restricted, uncertain, or synthetic. Faces are removed or blurred where they are not historically relevant, and retention rules are established for raw scans. Legal review separates public-domain material from material requiring permission. A provenance field is added to each reconstructed feature so that visitors can distinguish documented elements from interpretive additions.

A vertical slice is rebuilt using representative assets rather than the most detailed scans. Profiling identifies texture memory and shader complexity as the major browser constraints. The team adopts multiple levels of detail, compressed textures, asynchronous loading, and a simplified lighting model. For the headset, locomotion is redesigned with teleportation and controlled turning options, followed by user comfort testing.

Approach and Outcome

The final experience presents three visual states: documented reconstruction, scholarly interpretation, and clearly labelled imaginative atmosphere. A disclosure panel explains sources and uncertainty. A release checklist verifies rights, privacy treatment, performance, accessibility, provenance, and emergency rollback. Although the final product contains fewer decorative details than the original prototype, it is more credible, more portable, and easier to maintain.

Learning Points

The case demonstrates that governance is not an administrative addition to graphics work. It changes asset selection, interface design, rendering strategy, and public communication. Early classification and vertical-slice testing prevented expensive rework. Historical authenticity was treated as a managed level of evidence rather than as a visual effect. Technical optimization improved not only performance but also interpretability and accessibility.

Emerging Directions in Graphics

Several research directions are reshaping graphics engineering. Neural rendering uses learned representations or networks to synthesize, enhance, or reconstruct images and scenes. It can support denoising, super-resolution, view synthesis, material estimation, and real-time reconstruction, but it raises questions about generalization, hallucination, reproducibility, and hardware dependence.

Differentiable rendering makes aspects of the rendering process compatible with gradient-based optimization. This enables systems to infer scene parameters from images, optimize geometry or materials, and connect graphics with machine learning. Its limitations include imperfect physical models, non-differentiable operations, ambiguity in inverse problems, and sensitivity to data quality.

Neural and hybrid scene representations, including radiance fields, point-based methods, signed-distance representations, and learned compression, offer alternatives to traditional meshes and textures. Their practical adoption depends on editing capability, temporal stability, memory requirements, interoperability, and reliable export to production tools.

Spatial computing integrates physical and virtual environments through tracking, scene understanding, mixed reality, and three-dimensional interfaces. It requires careful treatment of spatial privacy, safety, occlusion, accessibility, and social presence. Digital twins connect graphics with live sensor data and simulation, creating valuable tools for engineering and operations but also increasing the consequences of inaccurate models.

Generative systems are becoming integrated into ideation, asset variation, animation assistance, material creation, and code generation. The strategic challenge is to use automation without losing provenance, skill development, human authorship, or accountability. Organizations should define approved use cases, review thresholds, disclosure rules, and retention of prompts, models, and source material.

Research evaluation should therefore consider not only benchmark quality. A mature assessment asks whether a method is reproducible, editable, scalable, interpretable, fair, secure, energy-conscious, and compatible with real production constraints. The most influential future systems may be those that connect strong visual performance with reliable governance.

Best-Practice Framework for Graphics Organizations

A graphics organization can use the following sequence as a practical governance cycle:

1. Define purpose and success criteria. State the users, decisions, visual goals, technical targets, and unacceptable uses.

2. Map assets and dependencies. Record origins, rights, formats, owners, sensitivity, and transformations.

3. Prototype a representative slice. Test quality, performance, interoperability, accessibility, and workflow before scaling production.

4. Choose technology deliberately. Compare build, buy, license, open-source, cloud, and partnership alternatives using lifecycle criteria.

5. Automate repeatable controls. Validate schemas, builds, dependencies, performance budgets, and reference renders.

6. Review human and social effects. Examine privacy, bias, safety, accessibility, cultural interpretation, and authenticity.

7. Release with traceability. Preserve versions, approvals, disclosures, rollback capability, and incident contacts.

8. Learn after deployment. Monitor defects, user feedback, energy use, security events, and changing legal or technical conditions.

This cycle is not a substitute for professional judgment. Its value lies in making important questions visible early enough to influence design.

Challenges and Common Mistakes

A frequent mistake is optimizing for visual fidelity before establishing requirements and platform constraints. Another is treating assets as disposable files rather than governed information. Teams also underestimate the cost of maintaining plugins, proprietary formats, render farms, model dependencies, and undocumented scripts.

Governance can fail through excessive bureaucracy as well as through neglect. Controls should be risk-based and proportionate. Requiring the same approval process for a private prototype and a public synthetic identity system wastes effort in one case and provides inadequate protection in the other.

Automation may create false confidence. A clean build does not prove that the visual result is meaningful, fair, historically accurate, or safe. Similarly, a detection score does not establish authenticity with certainty. Human review, domain expertise, and transparent uncertainty remain necessary.

Finally, organizations may focus on launch and ignore retirement. Assets, accounts, datasets, models, and spatial records need archival or deletion decisions. A system that cannot be safely updated, exported, or decommissioned represents a governance risk even if its initial release succeeds.

Chapter Summary

Graphics engineering treats visual computing as a complete socio-technical system. Effective projects begin with explicit objectives, measurable constraints, representative prototypes, milestones, resource plans, and risk ownership. Production pipelines require controlled asset flow, versioning, reproducible builds, standards, automated validation, and expert review.

Build-versus-buy and platform decisions should consider strategic value, lifecycle cost, interoperability, security, vendor dependence, and organizational capability. Governance must address provenance, intellectual property, privacy, bias, authenticity, accessibility, sustainability, and accountability. Synthetic media can provide substantial creative and educational value, but responsible use requires disclosure, consent, traceability, and human oversight.

Emerging methods such as neural rendering, differentiable graphics, spatial computing, digital twins, and generative systems expand the capabilities of graphics engineering while introducing new risks. The central professional principle is to align technical excellence with defensible purpose, responsible governance, and long-term maintainability.

Key Terms

Graphics engineering: The planning, implementation, integration, testing, and governance of graphics systems.

Asset: A reusable visual, geometric, material, animation, audio, or descriptive element.

Production pipeline: The organized flow of assets, code, data, review, and delivery activities.

Vertical slice: A small integrated sample demonstrating the intended quality and workflow.

Provenance: The recorded origin, ownership, transformation, and history of an artifact.

Interoperability: The ability of different systems to exchange and correctly interpret information.

Technical debt: Future cost caused by expedient decisions that reduce maintainability or quality.

Build-versus-buy analysis: Evaluation of whether a capability should be developed internally or acquired externally.

Total cost of ownership: The complete lifecycle cost of acquiring, operating, maintaining, and replacing a technology.

Asset register: A controlled inventory describing assets, versions, rights, dependencies, and status.

Regression testing: Testing that checks whether changes have damaged previously accepted behavior.

Color management: Controlled handling of color representations across devices, software, and file formats.

Synthetic media: Media substantially generated or altered using computational or machine-learning methods.

Neural rendering: Use of learned models to generate, reconstruct, enhance, or transform visual content.

Differentiable rendering: Rendering formulated to support gradient-based optimization of scene or image parameters.

Spatial computing: Computing that integrates digital content with three-dimensional physical or virtual space.

Digital twin: A computational representation of a physical system connected to relevant data or state.

Data minimization: Collecting and retaining only the data necessary for a specified purpose.

Authenticity: The degree to which the origin, alteration, and status of media are accurately communicated.

Lifecycle sustainability: Consideration of environmental, social, and operational effects from creation through retirement.

Review Questions

Short Answer Questions

1. What is the purpose of a vertical slice in a graphics project?

2. Define provenance and explain why it matters for synthetic media.

3. Name four criteria used in a build-versus-buy decision.

4. What is regression testing in a graphics pipeline?

5. Why can spatial scans create privacy concerns?

Long Answer Questions

1. Explain how a graphics production pipeline can be governed from pre-production through release.

2. Compare building, buying, licensing, and partnering as strategies for acquiring graphics capability.

3. Discuss the relationship among intellectual property, asset registers, and provenance.

4. Explain how sustainability should influence graphics architecture and production practice.

5. Analyze the opportunities and limitations of neural rendering and differentiable graphics.

Analytical / Application Questions

1. A studio is choosing between a proprietary engine and an open framework for a ten-year project. Construct a decision framework that includes cost, portability, skills, vendor risk, performance, and legal considerations.

2. Design a governance plan for a system that generates realistic avatars from user photographs. Address consent, privacy, bias testing, disclosure, security, and deletion.

3. A museum reconstruction is visually impressive but historically uncertain and too large for mobile devices. Propose a strategy that balances evidence, performance, accessibility, and public communication.

22. Multiple Choice Questions with Answers

Question 1

Which project artifact most effectively demonstrates whether a graphics pipeline works across several integrated stages?

A. A list of software licenses

B. A vertical slice

C. A final marketing image

D. A hardware procurement plan

Correct Answer: B. A vertical slice

Question 2

Which factor is most important when evaluating the long-term risk of adopting a proprietary graphics platform?

A. The color of its default interface

B. The number of sample scenes supplied

C. The ability to export data and avoid excessive vendor dependence

D. The popularity of its logo

Correct Answer: C. The ability to export data and avoid excessive vendor dependence

Question 3

What is the primary purpose of an asset register?

A. To increase polygon counts automatically

B. To record asset identity, rights, dependencies, versions, and status

C. To replace all human reviews

D. To guarantee that an asset is historically accurate

Correct Answer: B. To record asset identity, rights, dependencies, versions, and status

Question 4

A continuous-integration system renders approved reference scenes after every shader change. This is primarily an example of what practice?

A. Visual regression testing

B. Data minimization

C. Platform certification

D. Creative outsourcing

Correct Answer: A. Visual regression testing

Question 5

Which situation presents the clearest privacy risk in a spatial-computing application?

A. Using a compressed texture format

B. Recording a user’s room geometry and retaining it without a clear purpose

C. Reducing the number of scene lights

D. Using level-of-detail models

Correct Answer: B. Recording a user’s room geometry and retaining it without a clear purpose

Question 6

Why is technical accuracy alone insufficient for evaluating bias in a visual AI system?

A. Accuracy cannot be measured on any dataset

B. A system may perform unequally across affected groups or encode harmful assumptions

C. Bias exists only in manual artwork

D. Technical systems cannot generate images

Correct Answer: B. A system may perform unequally across affected groups or encode harmful assumptions

Question 7

Which safeguard best supports responsible use of synthetic media?

A. Concealing all information about how content was produced

B. Combining provenance, disclosure, consent, and human review

C. Relying only on a watermark detector

D. Removing all metadata before release

Correct Answer: B. Combining provenance, disclosure, consent, and human review

Question 8

Which decision most directly improves sustainability in a graphics application?

A. Using maximum resolution for every device

B. Disabling asset reuse

C. Applying adaptive quality and level-of-detail management

D. Duplicating all textures in storage

Correct Answer: C. Applying adaptive quality and level-of-detail management

Question 9

What distinguishes differentiable rendering from ordinary rendering?

A. It cannot produce images

B. It supports gradient-based optimization of scene or image parameters

C. It is limited to printed graphics

D. It eliminates the need for data

Correct Answer: B. It supports gradient-based optimization of scene or image parameters

Question 10

In the illustrative heritage case, why were reconstructed features labelled by evidential status?

A. To increase polygon density

B. To distinguish documented history from interpretation and imaginative additions

C. To avoid all user testing

D. To make licensing unnecessary

Correct Answer: B. To distinguish documented history from interpretation and imaginative additions