arxiv: 2604.04096 · v1 · submitted 2026-04-05 · 💻 cs.SE · cs.AI

Recognition: no theorem link

Toward a Sustainable Software Architecture Community: Evaluating ICSA's Environmental Impact

Mahyar T. Moghaddam , Mina Alipour , Torben Worm , Mikkel Baun Kj{\ae}rgaard

Authors on Pith no claims yet

Pith reviewed 2026-05-13 17:32 UTC · model grok-4.3

classification 💻 cs.SE cs.AI

keywords carbon footprintgenerative AIsoftware architectureconference sustainabilityICSAenvironmental auditresearch emissionssustainable computing

0 comments

The pith

A carbon audit measures emissions from generative AI in software architecture papers alongside the full footprint of the ICSA conference.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper conducts the first audit of carbon emissions linked to generative AI in software architecture research papers and to the operations of the ICSA conference. It calculates separate inventories for AI inference in accepted papers and for attendee travel, accommodation, and venue use at ICSA 2025. These numbers allow the community to see the scale of its environmental impact and to consider changes toward lower emissions. Sympathetic readers would care because software research increasingly relies on AI tools whose energy demands add to the climate burden already present in conference travel. The work aims to foster a more sustainable culture by providing concrete data for reflection and action.

Core claim

The authors establish two carbon footprints for the ICSA context: one exploratory calculation of emissions from generative AI tools used in preparing accepted papers, bounded by research artifacts, and another complete accounting of emissions from conference attendance and operations for the 2025 event, covering travel, lodging, food, energy, and materials. These inventories, though differing in completeness, together support discussions on how the community can reduce its overall environmental load through transparency and targeted efficiency measures.

What carries the argument

Two separate carbon inventories defined by distinct system boundaries, one for generative AI inference usage and one for conference attendance and operations.

Load-bearing premise

The audit depends on assumptions that the rates of generative AI usage in papers are typical and that standard emission factors accurately represent local conditions for travel and energy use.

What would settle it

Collecting actual usage data from authors on how often they used specific GenAI tools during paper writing or obtaining precise attendee travel records would test the accuracy of the reported emission totals.

Figures

Figures reproduced from arXiv: 2604.04096 by Mahyar T. Moghaddam, Mikkel Baun Kj{\ae}rgaard, Mina Alipour, Torben Worm.

**Figure 2.** Figure 2: Research methodology. III. RESEARCH METHODOLOGY A. Data Collection and Scope Papers Data: We analyzed the 108 accepted ICSA 2025 papers (28 main track, 80 companion) to ensure a manageable and replicable scope. Ethics and data protection are discussed in Section III-F. Activity Data: To estimate emissions from conference logistics, we collected organizer-provided data on 229 inperson attendees, their geog… view at source ↗

**Figure 3.** Figure 3: The comparison of worldwide regions participants (full conference) [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

read the original abstract

Generative AI (GenAI) tools are increasingly integrated into software architecture research, yet the environmental impact of their computational usage remains largely undocumented. This study presents the first systematic audit of the carbon footprint of both the digital footprint from GenAI usage in research papers, and the traditional footprint from conference activities within the context of the IEEE International Conference on Software Architecture (ICSA). We report two separate carbon inventories relevant to the software architecture research community: i) an exploratory estimate of the footprint of GenAI inference usage associated with accepted papers within a research-artifact boundary, and ii) the conference attendance and operations footprint of ICSA 2025 (travel, accommodation, catering, venue energy, and materials) within the conference time boundary. These two inventories, with different system boundaries and completeness, support transparency and community reflection. We discuss implications for sustainable software architecture, including recommendations for transparency, greener conference planning, and improved energy efficiency in GenAI operations. Our work supports a more climate-conscious research culture within the ICSA community and beyond

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

This paper runs a first-pass carbon audit on ICSA papers and the 2025 conference but the GenAI half depends on untested usage assumptions.

read the letter

The paper's core move is to produce two separate inventories for the software architecture community: an exploratory count of GenAI inference emissions tied to accepted papers, and a more conventional tally of travel, venue, and operations for ICSA 2025. It frames both as a way to start community reflection on sustainability rather than deliver definitive numbers. That framing is honest and the choice to treat the two inventories with different boundaries makes sense given the data limits. The conference side uses standard emission factors and covers the usual categories, so it is at least reproducible in principle. The GenAI side applies existing inference energy estimates to papers but relies on assumed representative usage rates across the accepted set without author surveys or sensitivity runs shown. That keeps the GenAI numbers illustrative. The work cites prior carbon-accounting literature appropriately and does not overclaim novelty beyond being the first such audit for this specific series. No load-bearing equations or fitted parameters appear, so circularity is not an issue. The main limitation is evidential depth on the GenAI inventory; without primary usage data the claim of a systematic audit is only partly met. Readers already working on green computing or conference sustainability will find the discussion useful as a prompt. Others will see it as a starting point that needs tighter methods before it influences practice. I would send this to peer review so referees can press on the assumption validation and boundary choices, but it does not yet look ready for a strong journal without revision.

Referee Report

1 major / 2 minor

Summary. The paper claims to present the first systematic audit of the carbon footprint of both GenAI usage in research papers and traditional conference activities for the IEEE International Conference on Software Architecture (ICSA). It reports two inventories: an exploratory estimate of the GenAI inference footprint associated with accepted papers and the full conference operations footprint for ICSA 2025 (travel, accommodation, catering, venue energy, and materials), followed by discussion of implications and recommendations for sustainable practices.

Significance. If the estimates can be substantiated with primary data and validation, the work could meaningfully advance climate-conscious practices in the software architecture community by highlighting both digital and physical impacts and prompting changes in conference planning and GenAI operations.

major comments (1)

[Abstract] Abstract: the claim of the 'first systematic audit' of both footprints is load-bearing on the GenAI inventory, which is described as exploratory and rests on assumed representative rates of AI tool usage across accepted papers plus accurate inference energy figures, with no primary data (e.g., author surveys), validation steps, or sensitivity analysis described to support the numbers.

minor comments (2)

The abstract would be strengthened by including at least headline numerical results and data sources so readers can immediately assess the scale of each inventory.
Add an explicit limitations subsection that quantifies uncertainty in the GenAI assumptions and discusses boundary choices for both inventories.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their constructive feedback on the abstract and the strength of our claims. We agree that the 'first systematic audit' phrasing requires qualification given the exploratory nature of the GenAI inventory and will revise the manuscript to address this while preserving the value of the conference operations audit.

read point-by-point responses

Referee: [Abstract] Abstract: the claim of the 'first systematic audit' of both footprints is load-bearing on the GenAI inventory, which is described as exploratory and rests on assumed representative rates of AI tool usage across accepted papers plus accurate inference energy figures, with no primary data (e.g., author surveys), validation steps, or sensitivity analysis described to support the numbers.

Authors: We agree that the GenAI inventory is exploratory and relies on literature-derived assumptions for usage rates and inference energy rather than primary data such as author surveys. The manuscript already labels this component as an 'exploratory estimate' and notes differing system boundaries. We will revise the abstract to qualify the overall contribution as combining an exploratory GenAI estimate with a more complete conference operations inventory, removing the unqualified 'first systematic audit' claim. We will also add a sensitivity analysis on key parameters (e.g., usage rates and energy figures) and expand the limitations section to discuss uncertainties and the absence of primary data collection. Primary data via surveys cannot be added in revision as it would require new empirical work outside the current study's scope; we will explicitly flag this as a direction for future research. revision: partial

Circularity Check

0 steps flagged

No circularity: exploratory audit relies on external factors and stated assumptions

full rationale

The paper presents two separate carbon inventories—an exploratory GenAI inference estimate and a conference operations footprint—using external emission factors, standard conversion values, and explicitly stated assumptions about usage rates. No equations, fitted parameters, predictions derived from prior results, or self-citations appear in the provided text that would reduce any claimed quantity to its own inputs by construction. The derivation chain consists of direct multiplication of activity data by independent coefficients; the result is therefore not equivalent to the inputs by definition. This is the normal non-circular outcome for an audit-style study.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

The central claim rests on standard life-cycle assessment emission factors for travel and energy plus unstated assumptions about average GenAI usage per paper; no new entities are introduced.

free parameters (2)

GenAI inference usage rate per paper
Exploratory estimate required to scale the digital footprint inventory
Conference-specific emission factors
Chosen values for travel, accommodation, catering, and venue energy

axioms (1)

domain assumption Standard carbon emission factors from life-cycle databases are accurate and transferable to this conference setting
Invoked to convert activity data into CO2-equivalent emissions for the conference inventory

pith-pipeline@v0.9.0 · 5495 in / 1217 out tokens · 74374 ms · 2026-05-13T17:32:16.946467+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

36 extracted references · 36 canonical work pages · 1 internal anchor

[1]

Generative ai for software architecture. applications, trends, challenges, and future directions,

M. Esposito, X. Li, S. Moreschini, N. Ahmad, T. Cerny, K. Vaid- hyanathan, V . Lenarduzzi, and D. Taibi, “Generative ai for software architecture. applications, trends, challenges, and future directions,” Applications, Trends, Challenges, and Future Directions, 2025

work page 2025
[2]

Energy and policy con- siderations for deep learning in nlp,

E. Strubell, A. Ganesh, and A. McCallum, “Energy and policy con- siderations for deep learning in nlp,” inProceedings of the 57th annual meeting of the assoc for computational linguistics, 2019, pp. 3645–3650

work page 2019
[3]

How hungry is ai? benchmarking energy, water, and carbon footprint of llm inference,

N. Jegham, M. Abdelatti, C. Y . Koh, L. Elmoubarki, and A. Hendawi, “How hungry is ai? benchmarking energy, water, and carbon footprint of llm inference,”arXiv preprint arXiv:2505.09598, 2025

work page arXiv 2025
[4]

The gentle singularity,

S. Altman, “The gentle singularity,”Sam Altman Blog, 2025, accessed: 2025-12-10. [Online]. Available: https://blog.samaltman.com/ the-gentle-singularity

work page 2025
[5]

Examining the carbon footprint of conferences with an emphasis on energy consumption and catering,

V . Mannheim and J. L. Avat ´o, “Examining the carbon footprint of conferences with an emphasis on energy consumption and catering,” Energies, vol. 18, no. 2, p. 321, 2025

work page 2025
[6]

How sustainable are sustainability conferences?–comprehensive life cycle assessment of an international conference series in europe,

S. Neugebauer, M. Bolz, R. Mankaa, and M. Traverso, “How sustainable are sustainability conferences?–comprehensive life cycle assessment of an international conference series in europe,”Journal of cleaner production, vol. 242, p. 118516, 2020

work page 2020
[7]

Generative artificial intelligence for software engineering—a research agenda,

A. Nguyen-Duc, B. Cabrero-Daniel, A. Przybylek, C. Arora, D. Khanna, T. Herda, U. Rafiq, J. Melegati, E. Guerra, K.-K. Kemellet al., “Generative artificial intelligence for software engineering—a research agenda,”Software: Practice and Experience, vol. 55, no. 11, pp. 1806– 1843, 2025

work page 2025
[8]

Large language models for software engineering: Sur- vey and open problems,

A. Fan, B. Gokkaya, M. Harman, M. Lyubarskiy, S. Sengupta, S. Yoo, and J. M. Zhang, “Large language models for software engineering: Sur- vey and open problems,” in2023 IEEE/ACM International Conference on Software Engineering: Future of Software Engineering (ICSE-FoSE). IEEE, 2023, pp. 31–53

work page 2023
[9]

Can llms generate architec- tural design decisions?-an exploratory empirical study,

R. Dhar, K. Vaidhyanathan, and V . Varma, “Can llms generate architec- tural design decisions?-an exploratory empirical study,” in2024 IEEE 21st International Conference on Software Architecture (ICSA). IEEE, 2024, pp. 79–89

work page 2024
[10]

Do large language models contain software architectural knowledge?: An exploratory case study with gpt,

M. Soliman and J. Keim, “Do large language models contain software architectural knowledge?: An exploratory case study with gpt,” in2025 IEEE 22nd International Conference on Software Architecture (ICSA). IEEE, 2025, pp. 13–24

work page 2025
[11]

Swiss cheese model for ai safety: A taxonomy and reference architecture for multi-layered guardrails of foundation model based agents,

M. Shamsujjoha, Q. Lu, D. Zhao, and L. Zhu, “Swiss cheese model for ai safety: A taxonomy and reference architecture for multi-layered guardrails of foundation model based agents,” in2025 IEEE 22nd Int Conference on Software Architecture (ICSA). IEEE, 2025, pp. 37–48

work page 2025
[12]

Improving clinical decision support: Architecture design of a multi-agent system based on an argument quality assessment ontology,

P. Liu and L. Xiao, “Improving clinical decision support: Architecture design of a multi-agent system based on an argument quality assessment ontology,” in2025 IEEE 22nd International Conference on Software Architecture (ICSA). IEEE, 2025, pp. 313–323

work page 2025
[13]

Enabling architec- ture traceability by llm-based architecture component name extraction,

D. Fuchß, H. Liu, T. Hey, J. Keim, and A. Koziolek, “Enabling architec- ture traceability by llm-based architecture component name extraction,” in2025 IEEE 22nd International Conference on Software Architecture (ICSA). IEEE, 2025, pp. 1–12

work page 2025
[14]

Llms for generation of architectural components: An exploratory empirical study in the server- less world,

S. Arun, M. Tedla, and K. Vaidhyanathan, “Llms for generation of architectural components: An exploratory empirical study in the server- less world,” in2025 IEEE 22nd International Conference on Software Architecture (ICSA). IEEE, 2025, pp. 25–36

work page 2025
[15]

Edgemlbalancer: A self-adaptive approach for dynamic model switching on resource-constrained edge devices,

A. Matathammal, K. Gupta, L. Lavanya, A. V . Halgatti, P. Gupta, and K. Vaidhyanathan, “Edgemlbalancer: A self-adaptive approach for dynamic model switching on resource-constrained edge devices,” in 2025 IEEE 22nd International Conference on Software Architecture Companion (ICSA-C). IEEE, 2025, pp. 543–552

work page 2025
[16]

Dynamic architectures leveraging ai agents and human-in-the-loop for data center management,

V . D. Sera and S. Kalra, “Dynamic architectures leveraging ai agents and human-in-the-loop for data center management,” in2025 IEEE 22nd International Conference on Software Architecture Companion (ICSA- C). IEEE, 2025, pp. 324–333

work page 2025
[17]

Visualizing ai energy use: Can consumption reminders promote social good mindsets?

E. Barnett, “Visualizing ai energy use: Can consumption reminders promote social good mindsets?” Ph.D. dissertation, Texas Christian University Fort Worth, Texas, 2025

work page 2025
[18]

How co2stly is chi? the carbon footprint of generative ai in hci research and what we should do about it,

N. Inie, J. Falk, and R. Selvan, “How co2stly is chi? the carbon footprint of generative ai in hci research and what we should do about it,” inProceedings of the 2025 CHI Conference on Human Factors in Computing Systems, 2025, pp. 1–29

work page 2025
[19]

Carbon Emissions and Large Neural Network Training

D. Patterson, J. Gonzalez, Q. Le, C. Liang, L.-M. Munguia, D. Rothchild, D. So, M. Texier, and J. Dean, “Carbon emissions and large neural network training,”arXiv preprint arXiv:2104.10350, 2021

work page internal anchor Pith review Pith/arXiv arXiv 2021
[20]

Trends in ai inference energy consumption: Beyond the performance-vs-parameter laws of deep learning,

R. Desislavov, F. Mart ´ınez-Plumed, and J. Hern´andez-Orallo, “Trends in ai inference energy consumption: Beyond the performance-vs-parameter laws of deep learning,”Sustainable Computing: Informatics and Sys- tems, vol. 38, p. 100857, 2023

work page 2023
[21]

Trend towards virtual and hybrid conferences may be an effective climate change mitigation strategy,

Y . Tao, D. Steckel, J. J. Kleme ˇs, and F. You, “Trend towards virtual and hybrid conferences may be an effective climate change mitigation strategy,”Nature communications, vol. 12, no. 1, p. 7324, 2021

work page 2021
[22]

Life cycle assessment of in-person, virtual, and hybrid academic conferences: New evidence and perspectives,

A. Cavallin Toscani, A. Atasu, L. N. Van Wassenhove, and A. Vinelli, “Life cycle assessment of in-person, virtual, and hybrid academic conferences: New evidence and perspectives,”Journal of Industrial Ecology, vol. 27, no. 6, pp. 1461–1475, 2023

work page 2023
[23]

Emission reduction potentials for academic conference travel,

S. Van Ewijk and P. Hoekman, “Emission reduction potentials for academic conference travel,”Journal of Industrial Ecology, vol. 25, no. 3, pp. 778–788, 2021

work page 2021
[24]

Hotel carbon measurement initiative (hcmi) methodology – version 2.0,

Sustainable Hospitality Alliance, “Hotel carbon measurement initiative (hcmi) methodology – version 2.0,” https://sustainablehospitalityalliance. org/wp-content/uploads/2020/02/HCMI-v2.0-introduction.pdf, 2020, ac- cessed: 2025-12-10

work page 2020
[25]

Re- viewing the carbon footprint analysis of hotels: Life cycle energy analysis (lcea) as a holistic method for carbon impact appraisal of tourist accommodation,

V . Filimonau, J. Dickinson, D. Robbins, and M. A. Huijbregts, “Re- viewing the carbon footprint analysis of hotels: Life cycle energy analysis (lcea) as a holistic method for carbon impact appraisal of tourist accommodation,”Journal of cleaner production, vol. 19, no. 17-18, pp. 1917–1930, 2011

work page 1917
[26]

Gptzero: Robust detection of llm-generated texts,

G. A. Adam, A. Cui, E. Thomas, E. Napier, N. Shmatko, J. Schnell, J. J. Tian, A. Dronavalli, E. Tian, and D. Lee, “Gptzero: Robust detection of llm-generated texts,” 2026. [Online]. Available: https://arxiv.org/abs/2602.13042

work page arXiv 2026
[27]

Ecologits: Evaluating the environmental impacts of generative ai,

S. Rinc ´e and A. Banse, “Ecologits: Evaluating the environmental impacts of generative ai,”Journal of Open Source Software, vol. 10, no. 111, p. 7471, 2025

work page 2025
[28]

Uk government ghg conversion factors for company reporting,

Department for Environment, Food & Rural Affairs (DEFRA), “Uk government ghg conversion factors for company reporting,” https://www.gov.uk/government/publications/ greenhouse-gas-reporting-conversion-factors-2023, 2023

work page 2023
[29]

Myclimate flight co2 emissions calculator,

Myclimate, “Myclimate flight co2 emissions calculator,” https://co2. myclimate.org/en/flight calculators/new, 2025, accessed: 2025-12-10

work page 2025
[30]

carbonlabel.org data,

CarbonLabel.org, “carbonlabel.org data,” https://carbonlabel.org/data, 2025, accessed: 2025-12-10

work page 2025
[31]

Climatiq emission factor, co2 (kg) per passenger-km, short-haul flight, economy class,

Climatiq, “Climatiq emission factor, co2 (kg) per passenger-km, short-haul flight, economy class,” https://www.climatiq.io/data/ emission-factor/31c8e342-a53e-4ae0-8fb7-dc1e69fa6965, 2025, accessed: 2025-12-10

work page 2025
[32]

Building energy audit, thermal comfort, and iaq assessment of a school building: A case study,

A. Merabtine, C. Maalouf, A. A. W. Hawila, N. Martaj, and G. Polidori, “Building energy audit, thermal comfort, and iaq assessment of a school building: A case study,”Building and Environment, vol. 145, pp. 62–76

work page
[33]

Electricity mix denmark,

NowTricity, “Electricity mix denmark,” https://www.nowtricity.com/ country/denmark/, 2025, accessed: 2025-12-10

work page 2025
[34]

Climate is served: Facts and figures on denmark’s food consumption and green- house gas emissions,

Danish Ministry of Food, Agriculture and Fisheries (FVM), “Climate is served: Facts and figures on denmark’s food consumption and green- house gas emissions,” https://en.fvm.dk/Media/638499759964791542/ Climate is Served.pdf, 2021, accessed: 2025-12-10

work page arXiv 2021
[35]

Den store klimadatabase,

Den Store Klimadatabase, “Den store klimadatabase,” https://denstoreklimadatabase.dk/en, 2025, accessed: 2025-12-10

work page 2025
[36]

Greenhouse gas inventory guidance,

U. EPA, “Greenhouse gas inventory guidance,” inIndirect Emissions from Events and Conferences. US EPA Center for Corporate Climate Leadership, 2018, pp. 2018–12

work page 2018