This draft is the version I used in the first supervision meeting (23 July, 2025) with Dr Hackl, my supervisor.
Competitive Algorithms, Cooperative Alternatives: Rethinking Space Telescope Scheduling Through Symbiogenetic Principles

Abstract

The Hubble and James Webb Space Telescopes cost hundreds of thousands of dollars per minute to operate, yet their observation schedules are determined by an algorithm called SPIKE that uses evolutionary computation to optimise telescope time. This research examines how SPIKE’s evolutionary algorithms (EA) function as gatekeepers of cosmic knowledge, embedding competitive Darwinian logic into decisions about what humanity observes in the universe. Through analysis of NASA’s scheduling data and technical documentation, I reveal how these algorithms may systematically favour certain types of research and institutions whilst marginalising others. Drawing on Lynn Margulis’s symbiogenetic theory—which emphasises cooperation over competition in evolution—I propose alternative scheduling principles that prioritise epistemic diversity and global equity. This matters because as algorithms increasingly govern resource allocation across society, understanding their embedded values becomes crucial for democratic knowledge production. The project offers both critique and constructive alternatives for algorithmic decision-making in science.

Rationale

This research examines how evolutionary algorithms (EA) shape scientific knowledge production through a case study of NASA’s SPIKE telescope scheduler. This matters for three interconnected reasons:

Why This Matters for Digital Society

As algorithms increasingly govern resource allocation—from bank loans to research funding—understanding their embedded values becomes crucial. SPIKE represents an extreme case of algorithmic governance: it decides which parts of the universe humanity observes, effectively determining what cosmic knowledge gets produced. This invisible algorithmic mediation shapes not just efficiency but fundamental questions about what counts as valuable knowledge.

Why This Matters for Science and Economy

Space telescopes cost hundreds of thousands of dollars per minute to operate. SPIKE’s evolutionary algorithm optimises this precious resource using “fitness functions” that encode specific assumptions about scientific value. My preliminary analysis suggests these algorithms may systematically favour certain types of research questions and institutions, potentially excluding astronomers from the Global South or those pursuing unconventional enquiries. When algorithms trained on competitive evolutionary logic determine scientific priorities, they risk reproducing existing inequalities in global knowledge production.

Why This Matters Now

The recent AI boom has accelerated algorithmic decision-making across all sectors. Yet we rarely examine the foundational logics—like Darwinian competition—built into these systems. This research provides a concrete example of how technical choices cascade into social outcomes. By understanding how evolutionary algorithms shape telescope access, we gain insights applicable to algorithmic resource allocation in healthcare, education, and public services.
The recent AI boom has accelerated algorithmic decision-making across all sectors, from social welfare to scientific research. Yet we rarely examine the foundational logics—like Darwinian competition—built into these systems. This research provides a concrete example of how technical choices cascade into social outcomes. Notably, SPIKE is not a historical artefact but an active scheduling system, still used by the Space Telescope Science Institute (STScI) to allocate observation time for both the Hubble Space Telescope and the James Webb Space Telescope. As of 2025, it remains the principal decision-making tool for managing telescope access, and is continuously updated to accommodate evolving mission constraints and scientific priorities. This makes it an ideal case study for examining how algorithmic architectures translate into epistemic gatekeeping. By understanding how evolutionary algorithms shape telescope access today, we gain insights applicable to algorithmic resource allocation in other sectors like healthcare, education, and public services. In addition, it is not merely a theoretical exercise, but a timely intervention into the ethics of algorithmic governance in live scientific infrastructures.

The Core Insight

Algorithms are not neutral tools but embed specific worldviews. When we use evolutionary algorithms to schedule telescopes, we’re not just optimising efficiency. In fact, we are encoding assumptions about competition, value, and what deserves to be seen. This research makes these hidden choices visible and asks: what if we designed algorithms based on cooperation rather than competition?

Literature Review

This research builds on two main bodies of scholarship: technical literature on evolutionary algorithms and critical studies of algorithmic knowledge production.

Evolutionary Algorithms in Astronomical Practice

Evolutionary algorithms (EA) form a family of optimisation techniques inspired by Darwinian natural selection, including Genetic Algorithms (GA), Evolution Strategies (ES), and Genetic Programming (GP) (Mitchell, 1998; De Jong, 2006). These algorithms operate through iterative cycles of selection, mutation, and reproduction, where “fitter” solutions survive to the next generation. In astronomy, EA have been deployed for tasks ranging from telescope scheduling to galaxy classification and signal detection.

SPIKE, developed in the 1980s for Hubble and later adapted for the James Webb Space Telescope, represents one of the earliest applications of EA to scientific resource allocation. The system treats observation proposals as competing organisms in an evolutionary landscape, with a fitness function determining which proposals best optimise telescope usage. Technical documentation reveals that fitness is calculated based on factors including target visibility windows, instrument configuration changes, and scientific priority scores—but how these weights are determined remains opaque.

Critical Perspectives on Algorithmic Knowledge Production

Critical algorithm studies have illuminated how technical systems encode social assumptions and reproduce power relations (Winner, 2017; Benjamin, 2019). Haraway’s (1988) notion of situated knowledges reminds us that algorithms, like any knowledge apparatus, are shaped by positionality and perspective. In the case of SPIKE, the underlying fitness functions do more than optimise—they quietly curate a vision of what scientific value ought to be.

While many recent discussions of AI bias centre on data or output disparities (Blodgett et al., 2020), deeper architectural logics often go unexamined. Barad (2007), drawing from quantum physics, offers a framework for understanding how measurement itself enacts the world. Algorithms like SPIKE, through their evaluation mechanisms, do not merely select observations—they participate in defining which phenomena are rendered visible, and which remain in the dark.

Against the grain of this competitive machinery, Margulis’s symbiogenetic theory offers a gentler provocation. Evolution, she argued, was not only a struggle for dominance but also a history of unlikely alliances of bacteria fusing, collaborating, creating entirely new forms of life. This spirit of reciprocity, though born in microbiology, holds imaginative potential for rethinking the architectures of algorithmic allocation.

These insights resonate with long-standing debates in the history and philosophy of science. Daston and Galison (2010) trace the emergence of objectivity alongside new scientific instruments, showing how ideals of neutrality co-evolve with technologies of vision. Astronomy, mediated through telescopes and algorithms, exemplifies this entanglement. The telescope doesn’t just extend our sight. It shapes what we come to know as the cosmos.

Bruno Latour’s laboratory ethnographies echo this perspective. Scientific facts, in his account, emerge through messy negotiations between humans and machines, numbers and narratives. Following this line, SPIKE is neither invisible nor impartial. It is a quiet actor in the production of cosmic truth, a gatekeeper of the stars. To study how it allocates telescope time is to study how our image of the universe is filtered, structured, and governed by code.

Case Study: How SPIKE Shapes Cosmic Knowledge

To understand the stakes of algorithmic mediation in science, consider how SPIKE actually works. When astronomers worldwide submit observation proposals to use Hubble or Webb, they enter a complex algorithmic competition they may not fully understand.

The Algorithmic Process

SPIKE receives thousands of proposals annually, each requesting specific observation windows, instruments, and exposure times. The algorithm treats these as a massive optimisation problem: how to pack the maximum “scientific value” into limited telescope time whilst respecting technical constraints like Earth occlusion and instrument availability.

The evolutionary algorithm works by:

1. Creating an initial population of possible schedules
2. Evaluating each schedule’s “fitness” based on multiple criteria
3. Selecting high-fitness schedules to “reproduce” with mutations
4. Iterating until an optimal schedule emerges

Potential Biases in Algorithmic Design

Based on how evolutionary algorithms typically function and SPIKE’s documented emphasis on efficiency, several concerns merit investigation:

• Efficiency optimisation: If the algorithm prioritises observations requiring minimal telescope movement and instrument changes, this could disadvantage innovative research requiring unique configurations
• Possible institutional effects: Systems that consider past performance might create feedback loops advantaging established institutions
• Geographic considerations: Time zones and ground station locations could create systematic access inequalities for certain regions

These potential biases warrant systematic investigation through data analysis.

Hypothetical Impacts

If these biases exist, a proposal to study an unusual celestial phenomenon requiring complex instrument configurations from a smaller university in the Global South could face multiple algorithmic disadvantages before any human evaluates its scientific merit. The algorithm’s efficiency optimisation could become a de facto gatekeeper determining whose questions about the universe get answered.

What Gets Lost

When optimisation becomes the sole criterion, we lose:

• Serendipitous discoveries from “inefficient” observations
• Diverse perspectives from researchers outside major institutions
• Unconventional approaches that don’t fit standard templates
• Equitable access to humanity’s shared scientific instruments

This isn’t just about fairness, but also about the kinds of knowledge we produce. When algorithms embed competitive logic into scientific infrastructure, they shape not just who can participate in science but what questions can be asked about our universe.

Methodology

This research employs Ruth Levitas’s “Utopia as Method” as an organising framework, using her triadic approach of excavation (archaeology), critique (ontology), and construction (architecture) to examine and reimagine algorithmic systems. Rather than relying on large-scale data science or predictive modelling, the study combines close textual analysis with lightweight digital humanities tools that allow for transparent, reflexive engagement with algorithmic infrastructures.

Phase 1: Excavation / Archaeology – Uncovering the Logics of SPIKE

In the archaeological mode, I examine the historical and infrastructural foundations of SPIKE, NASA’s scheduling algorithm, focusing on how its embedded logics shape scientific possibility. This involves:

• Document Analysis: Close reading of publicly available technical documents, algorithmic manuals, and user guides to trace how concepts such as ‘fitness’, ‘optimisation’, and ‘efficiency’ are operationalised in SPIKE.
• Simple Data Collection: Compiling basic data from NASA’s archives on telescope time allocations—including geographic distribution, institutional affiliations, and types of successful proposals.
• Visualisation: Using accessible tools such asTableau or Flourish, I will produce simple visualisations (e.g. bar charts, maps) to highlight observable patterns in time allocation and institutional representation.
• Timeline Construction: A dynamic timeline usingTimelineJS will situate SPIKE within broader developments in astronomical infrastructure and algorithmic governance.

This phase aims to render visible the historical sedimentation of values within SPIKE’s architecture and to contextualise its emergence within techno-scientific rationalities.

Phase 2: Critique / Ontology – Reading Algorithmic Rhetorics

Here, the project shifts to interrogating the underlying assumptions about knowledge, value, and fairness encoded in SPIKE’s design:

• Discourse Analysis: Applying critical reading to algorithmic documentation, with particular attention to the framing of optimisation and the presumed neutrality of computation.
• Textual Mining: UsingVoyant Tools, I will explore rhetorical frequencies and associations—e.g. how terms like “efficiency”, “scientific merit”, or “competition” appear and cluster within official texts.
• Conceptual Reflection: This analysis will be situated within critical debates on objectivity, governance, and the epistemological limits of algorithmic reasoning.

The aim is not to discredit SPIKE’s utility, but to examine how particular ontologies of science are privileged over others, often with subtle but material consequences for access and representation.

Phase 3: Construction / Architecture – Speculative Reconfiguration

The final phase explores how algorithmic systems might be otherwise. Drawing on symbiogenetic theory (Margulis, 1998) as an alternative evolutionary framework, I will outline speculative design directions that resist zero-sum competition:

• Principle Translation: Reimagining key algorithmic values through symbiotic metaphors—emphasising reciprocity, diversity, and interdependence.
• Alternative Metrics: Proposing new ‘fitness’ criteria such as epistemic diversity, geographical inclusion, methodological novelty, and the potential for serendipitous discovery.
• Prototype Framing: Rather than developing a technical prototype, I will sketch aconceptual model of an “endosymbiotic scheduler”—a thought experiment that signals how algorithmic architectures might evolve toward cooperation and plurality.

This methodology balances theoretical critique with modest empirical techniques suited to a humanities-trained researcher. Levitas’s framework provides a systematic way to move from diagnosis through critique to the construction of alternatives—essential for research that seeks not just to identify problems but to imagine different futures. By integrating digital tools that are both accessible and interpretively meaningful, the project foregrounds epistemological inquiry over technical mastery. The goal is to reflectively redesign: how might algorithms participate in shaping futures that are more just, inclusive, and generative?

Project Report Format: Comics-Format Dissertation plus an 8,000-word Analytical Essay

In considering the final project’s mode of presentation, I propose adopting a comics-based research format, an innovative approach successfully demonstrated by Nick Sousanis in Unflattening (Harvard University Press, 2015). This approach is selected for four compelling reasons:

Firstly, my personal proficiency and experience with comics as an expressive medium makes this choice both practical and advantageous. Previously, during my undergraduate studies, a comics project of mine attracted serious interest from a publisher, affirming my skill and the communicative potential of my artistic work.

Secondly, the comics-based format explicitly aligns with the Edinburgh Futures Institute’s emphasis on creativity-driven research and interdisciplinarity. By integrating textual scholarship with visual narrative, the dissertation can better reflect EFI’s core values: innovation, interdisciplinary inquiry, and creative experimentation.

Thirdly, the complex subject matter of astronomical observation and algorithmic knowledge production (as exemplified by NASA’s SPIKE scheduler) can significantly benefit from a visual storytelling format. Comics possess a unique capacity to make abstract, intricate theories and technical discussions tangible and engaging, thereby facilitating readers’ immersion into the narrative and encouraging deeper cognitive and affective connections to the research questions explored.

Lastly, adopting a multimodal output addresses practical constraints associated with word-count limits. While a traditional 15,000-word dissertation may restrict the detailed exposition my research demands, a comics-based dissertation, complemented by an 8,000-word analytical essay, would offer the necessary flexibility. Such a format allows for richer, more nuanced, and comprehensive exploration of key ideas, without sacrificing analytical depth or intellectual rigour.

In sum, a comics-format dissertation can leverage my strengths as a researcher-artist and also provide an ideal multimodal vehicle to convey my research effectively, creatively, and meaningfully.

Here are some excerpts from Sousanis’ work, Unflattening, the first comics-based research submitted as a doctoral dissertation at Teachers College, Columbia University.

https://www.youtube.com/watch?v=Ln7J10yn9iA

Timeline and Ethics

Project Timeline
Given the tight timeframe for the dissertation, I have designed an intensive five-week schedule for August, following an initial stage of literature review and supervisory consultation in late July. While the core analytical and theoretical work remains central to the dissertation, a visual component—in the form of a short research comic—has been introduced to reflect the visual and epistemic politics inherent in algorithmic telescope scheduling. This comic will serve as a speculative and critical device that complements and extends the written argument. The timeline below integrates both tracks:

Late July (22 July – 31 July)

• Conduct literature review on algorithmic visibility, feminist STS, and astronomy infrastructures.
• Deepen engagement with SPIKE’s underlying logic and document examples of how it prioritises or excludes telescope observations.
• Hold initial meeting with supervisor to clarify scope, methodology, and structure.
• Begin testing artistic workflow and style for visual component (e.g., layout, drawing tools, text–image balance).

Week 1 (1 – 7 August)

• Draft detailed outline of dissertation chapters (tentative: introduction, theoretical framework, case analysis, discussion).
• Develop an initial script and storyboard for the visual component, mapping how each concept or event will be expressed.
• Continue reviewing core readings and integrating citations.
• Begin sketching first 1–2 pages of the comic alongside the writing of the theoretical framework section.

Week 2 (8 – 14 August)

• Write a full draft of the theoretical framework and begin the case study section (SPIKE and HST scheduling politics).
• Draw and finalise comic pages 3–5, aligning visual narrative with theoretical insights.
• Begin shaping speculative elements in both textual and visual registers.

Week 3 (15 – 21 August)

• Complete case study and discussion section; revise theoretical framework based on new insights.
• Draft conclusion and review structure and transitions across the full essay.
• Continue drawing remaining pages of the visual component (pages 6–9).
• Begin assembling the final integrated format (text and visuals).

Week 4 (22 – 28 August)

• Finalise all drawings, captions, and annotations for the comic.
• Complete full integration of the visual and textual components, ensuring conceptual coherence.
• Review, proofread, and format entire dissertation for submission.

Week 5 / Buffer (29 – 31 August)

• Final edits, emergency adjustments, and formatting refinements.
• Submit dissertation before the final deadline.

This timeline assumes a clear division of tasks each week, with realistic workload distribution and fallback capacity. It also reflects my dual commitment: to rigorous critical inquiry and to experimenting with scholarly expression that engages both discursive and imagistic forms of knowledge.

Ethical Considerations

This research draws primarily on publicly available documentation and institutional archives, which minimises concerns regarding personal data or consent. However, the study is guided by the following ethical commitments:

1. Responsible Critique: While interrogating the epistemic assumptions embedded in SPIKE’s design, the research acknowledges the extraordinary scientific achievements the system has enabled.
2. Constructive Framing: Criticism is not levied in isolation but is accompanied by a speculative framework for more inclusive and pluralistic algorithmic futures.
3. Transparency: All interpretive procedures, sources, and theoretical frameworks will be clearly cited to ensure analytical accountability.
4. Engaged Dialogue: Should time permit interviews with astronomers or scheduling professionals, full ethical clearance will be sought, with attention to consent, confidentiality, and reflexivity.

The intention is not to discredit those involved in SPIKE’s development, but to open a reflective conversation about how algorithmic systems might better reflect values such as epistemic justice, accountability, and methodological diversity.

Expected Outcomes
This dissertation is expected to generate three primary outcomes. First, it will offer a critical analysis of algorithmic telescope scheduling, using SPIKE as a case study to examine how evolutionary algorithms participate in shaping what becomes visible—and what remains unseen—in astronomical knowledge production. This will foreground how specific logics of optimisation, efficiency, and quantifiability reconfigure epistemological access to the universe.

Second, it aims to intervene in broader debates around feminist epistemology and machine agency by drawing connections between algorithmic visibility regimes and colonial orders of seeing. By interrogating how Darwinian logics operate within seemingly neutral infrastructures, the project will contribute to ongoing critiques of instrumental rationality in AI design.

Third, the dissertation will include a short visual component in the form of a research comic. This comic is a speculative and reflexive method to visually unfold the entanglements between algorithms, space observation, and feminist critique. It is designed to offer an alternative epistemic lens through which to encounter the stakes of algorithmic mediation, particularly for audiences who engage with both textual and visual forms of knowledge.

Collectively, these outcomes aim to demonstrate how critical and speculative methodologies—rooted in feminist technoscience and digital humanities—can illuminate the political and ontological implications of AI systems in scientific discovery.

References

Barad, K. (2007). Meeting the universe halfway: Quantum physics and the entanglement of matter and meaning. Duke University Press.

Benjamin, R. (2019). Race after technology: Abolitionist tools for the new Jim Code. Polity.

Blodgett, S. L., Barocas, S., Daumé, H., III, & Wallach, H. (2020). Language (technology) is power: A critical survey of “bias” in NLP. Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics.
Daston, L., & Galison, P. (2010). Objectivity. Zone Books.

De Jong, K. A. (2006). Evolutionary computation: A unified approach. MIT Press.

Haraway, D. (1988). Situated knowledges: The science question in feminism and the privilege of partial perspective. Feminist Studies, 14(3), 575–599.
Latour, B. & Woolgar, S. (1979). Laboratory Life. Sage.

Margulis, L., & Sagan, D. (2023). Microcosmos: Four billion years of evolution from our microbial ancestors. University of California Press.

Mitchell, M. (1998). An introduction to genetic algorithms. MIT Press.
Sousanis, N. (2015). Unflattening. Harvard University Press.

Winner, L. (2017). Do artifacts have politics? In The whale and the reactor (pp. 19–39). University of Chicago Press.

Word Count: 3076 words (excluding title and references)

Aug 23, 2025