Naturalistic measure of social norms alignment

Kenneth Enevoldsen; Kristoffer Nielbo; M\'arton Kardos; Peter Bjerregaard Vahlstrup; Yevhen Kostiuk

arxiv: 2605.23420 · v1 · pith:BJJL6HK5new · submitted 2026-05-22 · 💻 cs.CL

Naturalistic measure of social norms alignment

Yevhen Kostiuk , Kenneth Enevoldsen , Peter Bjerregaard Vahlstrup , M\'arton Kardos , Kristoffer Nielbo This is my paper

Pith reviewed 2026-05-25 04:34 UTC · model grok-4.3

classification 💻 cs.CL

keywords social norms alignmentsolution matchinglanguage modelssocial dilemmasagreement metricsDanish datasetnaturalistic evaluationopen-ended responses

0 comments

The pith

Solution matching on open dilemmas measures how closely model responses align with panelist-derived social norms.

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

The paper introduces a framework to assess social norms alignment by matching solutions to free-form dilemmas instead of relying on multiple-choice or fixed-statement tests. This approach matters because everyday model use involves open conversations, where closed evaluations may not capture real alignment with shared expectations. The authors create a dataset of three thousand Danish social dilemmas, each paired with reference solutions from three panelists, and define two new metrics called stated agreement accuracy and explicit agreement accuracy. They apply the metrics to responses from several language models and humans in setups that mimic natural interactions. The results indicate that the metrics yield stable model rankings while also showing that agreement varies by dilemma topic.

Core claim

The solution-matching framework allows direct measurement of alignment between any pair of responses to social dilemmas by comparing them against culturally grounded reference solutions, producing consistent rankings across models and exposing topic-dependent differences in agreement levels.

What carries the argument

Solution matching between responses and panelist-derived references, operationalized through stated and explicit agreement accuracy metrics.

If this is right

The metrics support direct comparison of alignment between models, between models and humans, and among humans.
Agreement levels differ systematically across dilemma categories, reaching higher values for neighbor conflicts and shared living situations.
The same reference-based matching procedure can be applied to any new set of open-ended responses without requiring closed-form questions.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Extending the dataset construction process to additional languages would allow direct tests of whether alignment patterns reflect culture-specific norms.
The framework could be adapted to track changes in a single model's alignment after fine-tuning on culturally specific data.
If reference solutions are collected from more panelists per dilemma, the metrics might reduce sensitivity to individual judge variation.

Load-bearing premise

The solutions supplied by three panelists accurately represent the social norms that apply to each dilemma.

What would settle it

A larger survey of native Danish speakers that finds low agreement with the three panelists' reference solutions on a substantial fraction of dilemmas, or a replication in which the two metrics produce unstable model orderings across independent runs.

Figures

Figures reproduced from arXiv: 2605.23420 by Kenneth Enevoldsen, Kristoffer Nielbo, M\'arton Kardos, Peter Bjerregaard Vahlstrup, Yevhen Kostiuk.

**Figure 2.** Figure 2: Overview of the methodology. we extract solutions and their stances, match them via CMRs, and calculate the alignment metrics. Both steps were manually validated by human annotators; see subsection A.3 and subsection A.4 for details. Solutions Extraction For each dilemma d and a provided response r, we extract a set of advised (S + r ) and not advised (S − r ) solutions: (S + r , S− r ) = Extract(d, r) (1… view at source ↗

**Figure 3.** Figure 3: An overview of the topics contained in the corpus (left), and their distribution (right). We see that the [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

**Figure 4.** Figure 4: The data processing pipeline. Each episode is structured as follows. The host announces the dilemma, either reads it out loud or invites the author to provide details to the panel. After that, the panel discusses the potential solutions to the dilemma. After the discussion is concluded, the host turns to the next dilemma, usually having an advert or music break in between. At the end of the podcast, the h… view at source ↗

**Figure 5.** Figure 5: Results metrics: EAA, SAA, and their average [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗

**Figure 6.** Figure 6: Distributions on number of words in the mod [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗

**Figure 7.** Figure 7: Statistics of included entities in the open [PITH_FULL_IMAGE:figures/full_fig_p008_7.png] view at source ↗

**Figure 9.** Figure 9: Each panel corresponds to a fixed number [PITH_FULL_IMAGE:figures/full_fig_p014_9.png] view at source ↗

read the original abstract

Social norms reflect shared expectations on acceptable behavior. Measuring social norms alignment remains challenging, with existing approaches typically relying on artificial closed-form evaluations such as multiple-choice questionnaires or measuring agreement with predefined statements. In the context of this work, social norms alignment refers to measuring an agreement between solutions with respect to the social problem or dilemma. We propose a framework for measuring social norm alignment in naturalistic, free-form settings through solution matching. The framework enables us to measure alignment between any two dilemma responses e.g., LLMs to a human, LLMs to LLMs, or human to human. We introduce two metrics: stated and explicit agreement accuracy, and construct a dataset of 3k non-trivial social dilemmas in Danish. All dilemmas are assigned reference solutions derived from three panelists, who serve as culturally grounded judges. We evaluate the agreement of several LLMs and human responses in an interaction setup that resembles natural user-model conversations. Our results show that the proposed metrics produce consistent model rankings and reveal variation in agreement across different types of dilemmas, with higher agreement observed for topics such as neighbor conflicts and shared living situations. Overall, our work introduces a dataset and evaluation framework for studying culturally grounded social reasoning in naturalistic open-ended conversations.

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

New Danish dilemma dataset with solution-matching metrics for open-ended LLM evaluation, but three panelists as references is a thin foundation.

read the letter

The paper's core contribution is a 3k-item Danish dataset of social dilemmas plus a solution-matching method that scores free-form responses instead of forcing multiple-choice formats. It defines stated and explicit agreement accuracy and runs LLMs and humans in a conversational setup, reporting consistent model orderings and higher agreement on neighbor or shared-living topics. That shift toward naturalistic responses is the useful step; prior closed-form tests do miss real usage patterns, and releasing the data in Danish fills a language gap that English-only work leaves open. The results on topic variation are at least directionally interesting if the references hold up. The main weakness is exactly the reference construction. Three panelists generate the ground-truth solutions with no inter-rater numbers, majority-vote procedure, or external check against wider Danish opinion. On non-trivial dilemmas that can have legitimate disagreement, those three views could easily be idiosyncratic rather than representative. If the references drift, the reported consistency and topic differences become artifacts of the chosen judges rather than measures of norm alignment. The abstract gives no further detail on how dilemmas were sampled or how panelists were instructed, so the claim that the metrics track culturally grounded norms rests on an untested assumption. This work is aimed at researchers building culturally-aware evaluation suites or alignment benchmarks outside English. It is narrow in scope and the evidence on reference quality is light, but the dataset and the free-form matching idea are concrete enough that a serious referee should see it. I would send it for review with the expectation that the authors add reliability checks or external validation on the panel.

Referee Report

1 major / 2 minor

Summary. The paper proposes a framework for measuring social norms alignment via solution matching in naturalistic free-form responses to social dilemmas. It constructs a dataset of 3k non-trivial Danish dilemmas, each assigned reference solutions from three panelists as culturally grounded judges. Two metrics (stated agreement accuracy and explicit agreement accuracy) are introduced to quantify agreement between any pair of responses (LLM-LLM, LLM-human, human-human). LLMs and human responses are evaluated in an interaction setup resembling natural conversations; results indicate consistent model rankings across metrics and variation in agreement by dilemma type, with higher agreement on topics such as neighbor conflicts and shared living situations.

Significance. If the reference solutions validly represent shared norms, the work supplies a dataset and open-ended evaluation framework that improves on closed-form questionnaires for assessing culturally grounded social reasoning in LLMs. The reported consistency of rankings and topic-wise variation would constitute a useful empirical contribution for studying naturalistic alignment.

major comments (1)

[Dataset construction] Dataset construction (abstract and § on data collection): the claim that the metrics measure alignment with culturally grounded Danish social norms rests on the three panelists' solutions serving as stable, representative references for the 3k dilemmas. No inter-rater reliability, majority-vote procedure, or external validation is described; with only three judges the references may capture idiosyncratic views rather than shared expectations, especially on non-trivial dilemmas. This assumption is load-bearing: if it fails, the consistent model rankings and topic variation become artifacts of the chosen references rather than measurements of norm alignment.

minor comments (2)

[Abstract] The abstract states that the metrics produce 'consistent model rankings' but does not specify the exact set of models, the ranking procedure, or the statistical test used to establish consistency.
[Metrics] Notation for the two metrics (stated vs. explicit agreement accuracy) is introduced without an equation or pseudocode definition in the provided abstract; a formal definition would aid reproducibility.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address the major comment on dataset construction below.

read point-by-point responses

Referee: [Dataset construction] Dataset construction (abstract and § on data collection): the claim that the metrics measure alignment with culturally grounded Danish social norms rests on the three panelists' solutions serving as stable, representative references for the 3k dilemmas. No inter-rater reliability, majority-vote procedure, or external validation is described; with only three judges the references may capture idiosyncratic views rather than shared expectations, especially on non-trivial dilemmas. This assumption is load-bearing: if it fails, the consistent model rankings and topic variation become artifacts of the chosen references rather than measurements of norm alignment.

Authors: We agree that the stability and representativeness of the three panelists' reference solutions is a load-bearing assumption for interpreting the results as measurements of culturally grounded norms. The manuscript does not currently report inter-rater reliability or detail a majority-vote procedure. In the revised version we will add an analysis of agreement among the panelists (e.g., Fleiss' kappa and pairwise agreement rates) and describe the panelist selection and decision process in greater detail. External validation with a larger or independent group was outside the scope of the initial study; we will explicitly discuss this limitation and its implications for the reported model rankings and topic variation. revision: partial

Circularity Check

0 steps flagged

No significant circularity; metrics and references defined independently

full rationale

The paper defines its two agreement metrics (stated and explicit) directly from solution matching between responses and the panelist-derived references, without any equations that reduce the metrics to fitted parameters or self-referential quantities. The 3k-dilemma dataset and reference solutions are constructed as external inputs from three panelists rather than derived from the LLMs or the metrics themselves. No self-citations, uniqueness theorems, or ansatzes from prior author work are invoked to justify the central framework. The reported model rankings and topic-wise variation therefore rest on independent definitions and data rather than reducing to the inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the assumption that three panelists can supply culturally valid reference solutions; no free parameters or invented entities are described in the abstract.

axioms (1)

domain assumption Three panelists provide culturally grounded reference solutions that serve as valid ground truth for measuring alignment.
The metrics and evaluation depend on these references being accurate representations of social norms.

pith-pipeline@v0.9.0 · 5761 in / 1084 out tokens · 25934 ms · 2026-05-25T04:34:54.345716+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

20 extracted references · 20 canonical work pages · 3 internal anchors

[1]

Aligning AI With Shared Human Values

Aligning ai with shared human values.arXiv preprint arXiv:2008.02275. Matthew Honnibal, Ines Montani, Sofie Van Lan- deghem, and Adriane Boyd. 2020. spacy: Industrial- strength natural language processing in python. Kentaro Imajo, Masanori Hirano, Shuji Suzuki, and Hiroaki Mikami. 2025. A judge-free llm open-ended generation benchmark based on the distrib...

work page internal anchor Pith review Pith/arXiv arXiv 2008
[2]

Gemma 3 Technical Report

Gemma 3 technical report.Preprint, arXiv:2503.19786. Márton Kardos. 2026. SensTopic Documentation Page. Márton Kardos, Kenneth C. Enevoldsen, Jan Kostkan, Ross Deans Kristensen-McLachlan, and Roberta Rocca. 2025. Turftopic: Topic modelling with con- textual representations from sentence transformers. Journal of Open Source Software, 10(111):8183. Yevhen K...

work page internal anchor Pith review Pith/arXiv arXiv 2026
[3]

InFindings of the Association for Com- putational Linguistics: EMNLP 2025, pages 1–15, Suzhou, China

Automating alternative generation in decision- making. InFindings of the Association for Com- putational Linguistics: EMNLP 2025, pages 1–15, Suzhou, China. Association for Computational Lin- guistics. Woosuk Kwon, Zhuohan Li, Siyuan Zhuang, Ying Sheng, Lianmin Zheng, Cody Hao Yu, Joseph E. Gonzalez, Hao Zhang, and Ion Stoica. 2023. Effi- cient memory man...

work page 2025
[4]

Apache 2.0 licensed open-weight model

https://huggingface.co/mistralai/ Mistral-Small-3.2-24B-Instruct-2506 . Apache 2.0 licensed open-weight model. MistralAI. 2025b. mistralai/mistral-large-3-675b- instruct-2512. Ines Montani, Matthew Honnibal, Matthew Honni- bal, Sofie Van Landeghem, Adriane Boyd, Henning Peters, Paul O’Leary McCann, Jim Geovedi, Jim O’Regan, Maxim Samsonov, György Orosz, D...

work page
[5]

Joe Needham, Giles Edkins, Govind Pimpale, Henning Bartsch, and Marius Hobbhahn

explosion/spaCy: v3.5.0: New CLI commands, language updates, bug fixes and much more. Joe Needham, Giles Edkins, Govind Pimpale, Henning Bartsch, and Marius Hobbhahn. 2025. Large lan- guage models often know when they are being evalu- ated.arXiv preprint arXiv:2505.23836. OpenAI. 2025. gpt-oss-120b & gpt-oss-20b model card. Preprint, arXiv:2508.10925. Ord...

work page arXiv 2025
[6]

InThe Thirty-eighth Annual Conference on Neural Information Processing Systems

LLM evaluators recognize and favor their own generations. InThe Thirty-eighth Annual Conference on Neural Information Processing Systems. Alexis Plaquet and Hervé Bredin. 2023. Powerset multi- class cross entropy loss for neural speaker diarization. InProc. INTERSPEECH 2023. Jonathan Potter and Chloe Shaw. 2018. The virtues of naturalistic data. In Uwe Fl...

work page 2023
[7]

OpenAI GPT-5 System Card

SAGE Publications Ltd, London. Alec Radford, Jong Wook Kim, Tao Xu, Greg Brock- man, Christine McLeavey, and Ilya Sutskever. 2022. Robust speech recognition via large-scale weak su- pervision.arXiv preprint. Nils Reimers and Iryna Gurevych. 2019. Sentence-bert: Sentence embeddings using siamese bert-networks. InProceedings of the 2019 Conference on Empiri...

work page internal anchor Pith review Pith/arXiv arXiv 2022
[8]

jeg"- udsagn, f.eks

Undgå ballade og bare skrive under, 2) Skrive under kun hvis det præciseres, at min arv indsættes direkte på mine halvsøsk- endes børneopsparingskonti med klare re- gler for adgang (fx alder), eller 3) Spørge ind til, om de cirka 12,5% (min storebrors andel) virkelig er afgørende for børnenes økonomiske sikkerhed, når min far og hans kone tilsyneladende i...

work page 2019
[9]

På din side skifter I mellem mor og far

Sider-retfærdig (ingen to på stribe til samme side) - Juleaften: skiftevis din side og hans side. På din side skifter I mellem mor og far. - Eksempel: - 2019: Mor - 2020: Hans forældre - 2021: Far - 2022: Hans forældre - 2023: Mor - 2024: Hans forældre - Konsekvens: Hans forældre får juleaften hvert andet år; din mor og far hver fjerde. Afbødning: Fast li...

work page 2019
[10]

jul) - Juleaften roterer: Mor → Hans forældre → Far → gentag

Enheder-retfærdig (hver familieenhed hver 3. jul) - Juleaften roterer: Mor → Hans forældre → Far → gentag. - Eksempel: - 2019: Mor - 2020: Hans forældre - 2021: Far - 2022: Mor - 2023: Hans forældre - 2024: Far - Konsekvens: Alle tre enheder ser jer lige ofte, men din side får to på stribe (far → mor ved skiftet). Afbødning: Aftal at hans forældre fx alti...

work page 2019
[11]

I inviterer på skift: hans forældre, din mor, hans forældre, din far, osv

Hjemmejul (barn-venlig og konfliktned- trappende) - Juleaften altid hjemme hos jer. I inviterer på skift: hans forældre, din mor, hans forældre, din far, osv. - De, der ikke er med juleaften, får fast lillejuleaften/1. eller

work page
[12]

afgive” juleaften for altid – men I definerer, hvem der inviteres hvornår. Hvis du er ked af “kun hver fjerde jul

juledag. - Konsekvens: Minimalt re- jsekaos med baby, klare rammer. Ingen skal “afgive” juleaften for altid – men I definerer, hvem der inviteres hvornår. Hvis du er ked af “kun hver fjerde jul” hos din mor - Giv hende en fast, sikker plads hvert år (lillejuleaften eller 1. juledag), som I behandler som “lige så vigtig” som juleaften med faste traditioner...

work page
[13]

On your side, you alternate between mom and dad

Side-fair (no two in a row on the same side) - Christmas Eve: alternate your side and his side. On your side, you alternate between mom and dad. - Example: - 2019: Mom - 2020: His parents - 2021: Dad - 2022: His parents - 2023: Mom - 2024: 20 His parents - Consequence: His parents get Christmas Eve every other year; your mom and dad every fourth. Mitigati...

work page 2019
[14]

Unit-fair (each family unit every 3rd Christmas) - Christmas Eve rotates: Mother → His parents → Father → repeat. - Exam- ple: - 2019: Mother - 2020: His parents - 2021: Father - 2022: Mother - 2023: His parents - 2024: Father - Consequence: All three units see you equally often, but your side gets two in a row (father → mother when switching). Mitigation...

work page 2019
[15]

give up” Christmas Eve forever – but you define who is invited when. If you are upset about “only every fourth Christmas

Christmas at home (child-friendly and conflict-determining) - Christmas Eve al- ways at your home. You invite them in turns: his parents, your mother, his parents, your father, etc. - Those who are not with you on Christmas Eve will get a fixed date for Christmas Eve/Christmas Day or Christmas Day. - Consequence: Minimal travel chaos with the baby, clear ...

work page 2024
[16]

bestemme

**Del julen i 2025** (f.eks. juleaften hos jer, 1. juledag hos hans forældre, 2. juledag hos din mor), så alle får tid. 3. **Af- tal at holde jul hos jer selv hvert 3. år** (f.eks. 2026) og besøge familierne på andre tidspunkter (f.eks. 23. december eller 27. december). 4. **Lad barnet "bestemme" senere** – når det bliver ældre nok, kan I spørge, om det h...

work page 2025
[17]

Sig, at du forstår, hvorfor han føler, det er urimeligt

**Tag en snak med din kæreste uden at finde en løsning med det samme.** I stedet for at fokusere på modeller, så start med følelser. Sig, at du forstår, hvorfor han føler, det er urimeligt. Forklar ham, hvorfor det er så vigtigt for dig at se både din mor og far. Tal om, hvad der betyder allermest for jer hver især – er det juleaften takken, gud- stjenest...

work page
[18]

Okay, min idé var ikke fair. Hvad siger du til en af disse modeller i stedet?

**Præsenter de nye modeller som fælles bud.** Sig: "Okay, min idé var ikke fair. Hvad siger du til en af disse modeller i stedet?" Ved at præsentere dem sammen, viser du, at I er et team, der nu skal finde en løsning sammen. 3. **Vælg en model og prøv den af.** I er ikke låst fast for evigt. Sig: "Lad os prøve model 1 de første par år. Når barnet bliver s...

work page
[19]

every third year with each family unit

**Præsenter en samlet front.** Når I har besluttet jer, skal I fortælle det til jeres fam- ilier som en fælles, ubrydelig aftale. "Mor og far, vi har besluttet os for at gøre det så- dan her fremover, fordi det er det eneste, der 27 føles fair for både [kærestens navn] og mig. Vi glæder os utroligt meget til at fejre jul med jer på [den aftalte dag]." Det...

work page
[20]

stroke" of hair? 2. **Present the new mod- els as a joint bid.** Say:

**Have a talk with your boyfriend with- out finding a solution right away.** Instead of focusing on models, start with feelings. Say that you understand why he feels it is unreasonable. Explain to him why it is so important for you to see both your mother and father. Talk about what means the most to each of you – is it the Christmas Eve thanksgiving, the...

work page 2026

[1] [1]

Aligning AI With Shared Human Values

Aligning ai with shared human values.arXiv preprint arXiv:2008.02275. Matthew Honnibal, Ines Montani, Sofie Van Lan- deghem, and Adriane Boyd. 2020. spacy: Industrial- strength natural language processing in python. Kentaro Imajo, Masanori Hirano, Shuji Suzuki, and Hiroaki Mikami. 2025. A judge-free llm open-ended generation benchmark based on the distrib...

work page internal anchor Pith review Pith/arXiv arXiv 2008

[2] [2]

Gemma 3 Technical Report

Gemma 3 technical report.Preprint, arXiv:2503.19786. Márton Kardos. 2026. SensTopic Documentation Page. Márton Kardos, Kenneth C. Enevoldsen, Jan Kostkan, Ross Deans Kristensen-McLachlan, and Roberta Rocca. 2025. Turftopic: Topic modelling with con- textual representations from sentence transformers. Journal of Open Source Software, 10(111):8183. Yevhen K...

work page internal anchor Pith review Pith/arXiv arXiv 2026

[3] [3]

InFindings of the Association for Com- putational Linguistics: EMNLP 2025, pages 1–15, Suzhou, China

Automating alternative generation in decision- making. InFindings of the Association for Com- putational Linguistics: EMNLP 2025, pages 1–15, Suzhou, China. Association for Computational Lin- guistics. Woosuk Kwon, Zhuohan Li, Siyuan Zhuang, Ying Sheng, Lianmin Zheng, Cody Hao Yu, Joseph E. Gonzalez, Hao Zhang, and Ion Stoica. 2023. Effi- cient memory man...

work page 2025

[4] [4]

Apache 2.0 licensed open-weight model

https://huggingface.co/mistralai/ Mistral-Small-3.2-24B-Instruct-2506 . Apache 2.0 licensed open-weight model. MistralAI. 2025b. mistralai/mistral-large-3-675b- instruct-2512. Ines Montani, Matthew Honnibal, Matthew Honni- bal, Sofie Van Landeghem, Adriane Boyd, Henning Peters, Paul O’Leary McCann, Jim Geovedi, Jim O’Regan, Maxim Samsonov, György Orosz, D...

work page

[5] [5]

Joe Needham, Giles Edkins, Govind Pimpale, Henning Bartsch, and Marius Hobbhahn

explosion/spaCy: v3.5.0: New CLI commands, language updates, bug fixes and much more. Joe Needham, Giles Edkins, Govind Pimpale, Henning Bartsch, and Marius Hobbhahn. 2025. Large lan- guage models often know when they are being evalu- ated.arXiv preprint arXiv:2505.23836. OpenAI. 2025. gpt-oss-120b & gpt-oss-20b model card. Preprint, arXiv:2508.10925. Ord...

work page arXiv 2025

[6] [6]

InThe Thirty-eighth Annual Conference on Neural Information Processing Systems

LLM evaluators recognize and favor their own generations. InThe Thirty-eighth Annual Conference on Neural Information Processing Systems. Alexis Plaquet and Hervé Bredin. 2023. Powerset multi- class cross entropy loss for neural speaker diarization. InProc. INTERSPEECH 2023. Jonathan Potter and Chloe Shaw. 2018. The virtues of naturalistic data. In Uwe Fl...

work page 2023

[7] [7]

OpenAI GPT-5 System Card

SAGE Publications Ltd, London. Alec Radford, Jong Wook Kim, Tao Xu, Greg Brock- man, Christine McLeavey, and Ilya Sutskever. 2022. Robust speech recognition via large-scale weak su- pervision.arXiv preprint. Nils Reimers and Iryna Gurevych. 2019. Sentence-bert: Sentence embeddings using siamese bert-networks. InProceedings of the 2019 Conference on Empiri...

work page internal anchor Pith review Pith/arXiv arXiv 2022

[8] [8]

jeg"- udsagn, f.eks

Undgå ballade og bare skrive under, 2) Skrive under kun hvis det præciseres, at min arv indsættes direkte på mine halvsøsk- endes børneopsparingskonti med klare re- gler for adgang (fx alder), eller 3) Spørge ind til, om de cirka 12,5% (min storebrors andel) virkelig er afgørende for børnenes økonomiske sikkerhed, når min far og hans kone tilsyneladende i...

work page 2019

[9] [9]

På din side skifter I mellem mor og far

Sider-retfærdig (ingen to på stribe til samme side) - Juleaften: skiftevis din side og hans side. På din side skifter I mellem mor og far. - Eksempel: - 2019: Mor - 2020: Hans forældre - 2021: Far - 2022: Hans forældre - 2023: Mor - 2024: Hans forældre - Konsekvens: Hans forældre får juleaften hvert andet år; din mor og far hver fjerde. Afbødning: Fast li...

work page 2019

[10] [10]

jul) - Juleaften roterer: Mor → Hans forældre → Far → gentag

Enheder-retfærdig (hver familieenhed hver 3. jul) - Juleaften roterer: Mor → Hans forældre → Far → gentag. - Eksempel: - 2019: Mor - 2020: Hans forældre - 2021: Far - 2022: Mor - 2023: Hans forældre - 2024: Far - Konsekvens: Alle tre enheder ser jer lige ofte, men din side får to på stribe (far → mor ved skiftet). Afbødning: Aftal at hans forældre fx alti...

work page 2019

[11] [11]

I inviterer på skift: hans forældre, din mor, hans forældre, din far, osv

Hjemmejul (barn-venlig og konfliktned- trappende) - Juleaften altid hjemme hos jer. I inviterer på skift: hans forældre, din mor, hans forældre, din far, osv. - De, der ikke er med juleaften, får fast lillejuleaften/1. eller

work page

[12] [12]

afgive” juleaften for altid – men I definerer, hvem der inviteres hvornår. Hvis du er ked af “kun hver fjerde jul

juledag. - Konsekvens: Minimalt re- jsekaos med baby, klare rammer. Ingen skal “afgive” juleaften for altid – men I definerer, hvem der inviteres hvornår. Hvis du er ked af “kun hver fjerde jul” hos din mor - Giv hende en fast, sikker plads hvert år (lillejuleaften eller 1. juledag), som I behandler som “lige så vigtig” som juleaften med faste traditioner...

work page

[13] [13]

On your side, you alternate between mom and dad

Side-fair (no two in a row on the same side) - Christmas Eve: alternate your side and his side. On your side, you alternate between mom and dad. - Example: - 2019: Mom - 2020: His parents - 2021: Dad - 2022: His parents - 2023: Mom - 2024: 20 His parents - Consequence: His parents get Christmas Eve every other year; your mom and dad every fourth. Mitigati...

work page 2019

[14] [14]

Unit-fair (each family unit every 3rd Christmas) - Christmas Eve rotates: Mother → His parents → Father → repeat. - Exam- ple: - 2019: Mother - 2020: His parents - 2021: Father - 2022: Mother - 2023: His parents - 2024: Father - Consequence: All three units see you equally often, but your side gets two in a row (father → mother when switching). Mitigation...

work page 2019

[15] [15]

give up” Christmas Eve forever – but you define who is invited when. If you are upset about “only every fourth Christmas

Christmas at home (child-friendly and conflict-determining) - Christmas Eve al- ways at your home. You invite them in turns: his parents, your mother, his parents, your father, etc. - Those who are not with you on Christmas Eve will get a fixed date for Christmas Eve/Christmas Day or Christmas Day. - Consequence: Minimal travel chaos with the baby, clear ...

work page 2024

[16] [16]

bestemme

**Del julen i 2025** (f.eks. juleaften hos jer, 1. juledag hos hans forældre, 2. juledag hos din mor), så alle får tid. 3. **Af- tal at holde jul hos jer selv hvert 3. år** (f.eks. 2026) og besøge familierne på andre tidspunkter (f.eks. 23. december eller 27. december). 4. **Lad barnet "bestemme" senere** – når det bliver ældre nok, kan I spørge, om det h...

work page 2025

[17] [17]

Sig, at du forstår, hvorfor han føler, det er urimeligt

**Tag en snak med din kæreste uden at finde en løsning med det samme.** I stedet for at fokusere på modeller, så start med følelser. Sig, at du forstår, hvorfor han føler, det er urimeligt. Forklar ham, hvorfor det er så vigtigt for dig at se både din mor og far. Tal om, hvad der betyder allermest for jer hver især – er det juleaften takken, gud- stjenest...

work page

[18] [18]

Okay, min idé var ikke fair. Hvad siger du til en af disse modeller i stedet?

**Præsenter de nye modeller som fælles bud.** Sig: "Okay, min idé var ikke fair. Hvad siger du til en af disse modeller i stedet?" Ved at præsentere dem sammen, viser du, at I er et team, der nu skal finde en løsning sammen. 3. **Vælg en model og prøv den af.** I er ikke låst fast for evigt. Sig: "Lad os prøve model 1 de første par år. Når barnet bliver s...

work page

[19] [19]

every third year with each family unit

**Præsenter en samlet front.** Når I har besluttet jer, skal I fortælle det til jeres fam- ilier som en fælles, ubrydelig aftale. "Mor og far, vi har besluttet os for at gøre det så- dan her fremover, fordi det er det eneste, der 27 føles fair for både [kærestens navn] og mig. Vi glæder os utroligt meget til at fejre jul med jer på [den aftalte dag]." Det...

work page

[20] [20]

stroke" of hair? 2. **Present the new mod- els as a joint bid.** Say:

**Have a talk with your boyfriend with- out finding a solution right away.** Instead of focusing on models, start with feelings. Say that you understand why he feels it is unreasonable. Explain to him why it is so important for you to see both your mother and father. Talk about what means the most to each of you – is it the Christmas Eve thanksgiving, the...

work page 2026