Pith. sign in

REVIEW 5 major objections 5 minor 43 references

Fairness and Efficiency in Human-Agent Teams: An Iterative Algorithm Design Approach

T0 review · 5 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper claims that a task-allocation agent balancing efficiency with an equity-based fairness metric—equal ratios of preference-weighted outcomes to capabilities—is perceived as fairer and preferred over an efficiency-only agent by…

desk verdict An honest iterative-design paper whose abstract overclaims; the fair-equity metric is a useful contribution but the evidence only supports the Mixed team type, not 'various team types' or 'preferred.' read the letter →

arxiv 2505.16171 v1 pith:6WJXHGUT submitted 2025-05-22 cs.HC

Mai Lee Chang , Kim Baraka , Greg Trafton , Zach Lalu Vazhekatt , Andrea Lockerd Thomaz This is my paper · ORCID
classification cs.HC
keywords human-agentteamsfairtaskallocationfair-equitymetricfirst-personpointofviewefficiency-fairnesstrade-offiterativealgorithmdesignpreferenceshuman-robotinteraction
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to establish that a robot teammate allocating tasks to two humans can be both efficient and fair by optimizing a weighted reward that includes an equity-based fairness metric, and that people perceive such an agent as fairer and prefer it over an efficiency-only agent. The authors argue that the existing equality-of-capability metric, which fits a supervisor's third-person view, fails from the first-person view of the teammate doing the work. Their proposed fair-equity metric compares each person's preference-weighted outcomes to their capabilities, and the Fair-Efficient Algorithm (FEA) maximizes $\lambda E(s) + (1-\lambda)(1-|FE(s)|')$ with $\lambda = 0.70$. Simulation and small user studies lead them to conclude that FEA is perceived as fairer in Mixed team types, where the most capable member's preferences run against their capabilities, and that this fairness can be gained without a large efficiency loss. If the claim holds, fairness and efficiency need not trade off in human-agent task allocation, and fairness judgments should be treated as point-of-view dependent.

What carries the argument

The load-bearing object is the fair-equity metric $FE(s)$ (Eq. 9), defined as the difference between two teammates' outcome/input ratios: outcomes are $\sum_j p_{H_i,j} T_{H_i,j}(s)$ (tasks weighted by preference) and inputs are $\sum_j c_{H_i,j}$ (summed capabilities). This metric is normalized by min-max scaling (Eq. 11) using data from the first user study, and is inserted into the FEA reward (Eq. 10), a convex combination $\lambda E(s) + (1-\lambda)(1-|FE(s)|')$ with $\lambda=0.70$. The metric carries the argument because it turns Adams' equity theory into a per-state score that a policy-iteration MDP solver can optimize, and it also defines the paper's team taxonomy: Mixed, Twins, and Negative team types, distinguished by whether each member's capabilities and preferences are positively or negatively correlated.

What would settle it

Run the Mixed team type with enough participants for statistical tests and compare FEA against the Efficient algorithm on the preference-fairness and overall-fairness items: if participants do not rate FEA significantly fairer, or if they would not choose it for future tasks, the central claim is refuted. A sharper check is to vary the equity ratio while holding efficiency constant: if fairness ratings do not move with $FE(s)$, the metric is not what people are responding to.

Watch

Extended reading notes

Core claim

The central discovery, stated on the paper's own terms, is that fairness in human-agent task allocation is better captured by equity than by equality of capability. The paper defines fair-equity as the difference between two teammates' ratios of outcomes to inputs, where outcomes are preference-weighted task completions and inputs are summed capability coefficients (Eq. 9). It embeds this metric in the reward function $R(s)=\lambda E(s)+(1-\lambda)(1-|FE(s)|')$, with $\lambda=0.70$, and shows through simulation and two user studies that this Fair-Efficient Algorithm gives the less-capable teammate more rounds on their preferred and capable tasks, while the most-capable teammate also gets more preferred-task time than under an efficiency-only policy. The paper concludes that an equity-balancing agent will be perceived as fairer and preferred by human teammates, and that the perception of fairness depends on the person's point of view.

Load-bearing premise

The load-bearing premise is that a person's sense of fairness tracks the equity ratio of preference-weighted outcomes to capabilities, but the paper checks that correspondence with only seven participants in a single team type and normalizes the metric using an earlier six-participant sample.

Editorial extensions

If this is right

  • In Mixed teams, FEA shifts allocations so the less-capable teammate receives some rounds on their most capable and preferred tasks, while the most-capable teammate gains about 50% more rounds on preferred tasks than under the efficiency-only policy.
  • In Twins teams, where each member's capabilities and preferences align, FEA and the Efficient algorithm allocate almost identically, so adding fairness costs little when preferences already track capability.
  • The equality-of-capability metric does not transfer from a third-person to a first-person point of view: the most capable team members experienced equalizing allocations as unfair.
  • Fairness and efficiency are not necessarily in conflict in human-agent teamwork, since FEA achieved comparable efficiency for some teams while improving perceived fairness.
  • Transparency about what the allocating robot considers can raise fairness ratings, so the perceived fairness of FEA depends partly on participants having an accurate model of the algorithm.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial inference: the pairwise equity ratio suggests a natural generalization to $n>2$ teammates by comparing each member's outcome/input ratio to a team-level reference, but the paper only implements pairwise differences, so this extension is untested.
  • Editorial inference: because $\lambda=0.70$ was fixed in advance, the efficiency-fairness trade-off is left as a policy lever; future designs could tune $\lambda$ per team type or adapt it online when the less-capable teammate's preference gap is large.
  • Editorial inference: the point-of-view finding implies that fairness is stakeholder-relative, so an agent reasoning about fairness may have to decide whose point of view to optimize, a question the paper leaves open.
  • Editorial inference: a direct testable extension would run FEA in the Negative team type (both members' preferences negatively correlated with capabilities), where the equity metric's allocations and fairness ratings could differ from the Mixed team results.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 5 minor

Summary. This paper describes an iterative design process for task-allocation algorithms in a single-agent/two-human team. The authors define a fairness metric F based on capability and preference equality (Eqs. 1-2), an Efficient baseline (Eq. 4), and a Fair algorithm (Eq. 5). They evaluate these in simulation and in a six-person user study, then introduce a fair-equity metric FE (Eq. 9) and the Fair-Efficient Algorithm FEA (Eq. 10), with the min-max scaling in Eq. (11). FEA is evaluated in simulations across Mixed and Twins team types (Sec. 7.1) and in a seven-person user study of the Mixed team type (Sec. 7.2). The paper claims that an agent that balances efficiency and fairness based on equity is perceived as fairer and is preferred by human teammates in various human-agent team types.

Significance. If the central claim were supported, the fair-equity metric and FEA would be a useful contribution to human-agent task allocation, extending capability-based fairness to first-person perspectives and task preferences. The iterative simulation-plus-small-user-study method is a legitimate HCI design approach, and the authors are commendably transparent about their small samples and the exploratory nature of the work. However, the evidence is considerably weaker than the abstract suggests. The simulation 'predictions' of fairness are partly circular, the human validation has n=6 and n=7 with no inferential statistics, and only one team type was human-tested. As a design exploration the paper has value, but as a demonstration of generalizable fairness and efficiency benefits it would need substantial revision or additional evidence.

major comments (5)
  1. [Abstract; §7.1.2, §7.1.3, §7.2.1] The abstract's 'various human-agent team types' is not supported by the evidence. Section 7.1.2 reports that in the Twins team type the Efficient and FEA algorithms make identical allocations, so no fairness or preference difference is expected there. Section 7.1.3 explicitly limits FEA's benefit to the Mixed team type, and the only human evaluation (Sec. 7.2) is of the Mixed type with n=7. The 'preferred' claim is even weaker: Section 7.2.1 reports that two of four participants chose the Efficient robot and two chose FEA on both the 'better job' and 'would work with again' questions. The data, taken at face value, support at most 'in the Mixed team type FEA was rated fairer on the preference subscale,' not 'various team types' or 'preferred.'
  2. [§5.2.2, Fig. 3; §6.2, Eq. (11)] The simulation 'prediction' of fairness in Fig. 3 is not an independent test. For the Fair algorithm, the predicted fairness is the algorithm's own reward function 1-F(s) from Eq. (5) scaled to a 1-7 rating; because the algorithm maximizes exactly that quantity, a predicted rating of 7 is a restatement of the optimization objective. The FEA predictions in Fig. 5 inherit the same issue through Eq. (11), whose min-max constants are computed from the first user-study data (Sec. 6.2) and then reused for the second user study. Consequently the simulation-vs-human comparisons do not provide independent validation of the fairness metrics.
  3. [§5.2, §7.2] The human validation is underpowered and no inferential statistics are reported. Section 7.2 states n=7 and 'we do not perform statistical tests'; Section 5.2 has n=6 and the same statement. The headline differences (e.g., the preference-rating gap 5.14 vs 3.71 in Fig. 5) are reported as means with no variance, no paired comparisons, and no effect sizes. With n=7 and multiple Likert items, these means cannot support a conclusion that FEA is 'perceived as fairer' beyond a tentative trend. At minimum, appropriate paired tests or confidence intervals should be reported, and the paper's claims should be correspondingly hedged.
  4. [§6.2, Eq. (11)] The min-max normalization in Eq. (11) is fitted to the first user-study sample and then used to evaluate FEA on a different small sample, which is a form of data leakage. Because FEA's optimal policy depends on these constants, the reported simulation results are not robust to the scaling choice. The paper should either provide a principled, data-independent justification for the normalization constants or report sensitivity analyses over them.
  5. [§6.1, Eq. (9)] The fair-equity metric assumes that Adams' equity theory applies to human-agent task allocation, but the paper provides no independent evidence that the specific operationalization in Eq. (9) — preference-weighted task counts as outcomes and summed capabilities as inputs — corresponds to first-person fairness judgments. This construct-validity gap is load-bearing because the main result concerns perceived fairness. The metric should be validated separately from its use in the reward function, or the paper should clearly identify this assumption as an open question.
minor comments (5)
  1. [§1, Research Questions] The first research question says 'extend prior fairness metrics that has been shown to align with people's perception of fairness from a first-person POV to a third-person POV,' but the paper's actual direction (see §2.1 and §5.3) is extending third-person capability fairness to a first-person POV. The POV order appears to be reversed and should be corrected.
  2. [§8.1] The sentence 'Prior work's assessments were from a first-person POV and our work is from a third-person POV' appears to invert the paper's own characterization: prior work [7] is described in §2.1 as third-person, and this paper is described as first-person. Please correct this inconsistency.
  3. [§6.3 and §7.2.1] There are minor typographical errors: 'booth' should be 'both' in §6.3, and 'would like to to work' has a doubled 'to' in §7.2.1.
  4. [Figs. 3 and 5] The figures report means without any measure of spread or individual data points, which is difficult to interpret for n=6 and n=7. Consider boxplots or scatterplots, and label the predicted-fairness curves as the algorithms' own reward values rather than independent predictions.
  5. [§5.1] The fair-gap threshold of 0.60 is introduced without justification. The paper should explain how this threshold was chosen and whether the results are sensitive to it.

Circularity Check

2 steps flagged · score 5.0 of 10

Simulation-side 'predicted fairness' is the algorithm's own reward scaled to a 1-7 rating, so the simulation is not an independent test of the fairness metric; the human ratings are external but are small-n, single-team-type, and include a 2-2 preference tie.

  1. self definitional [Section 4.1, Eqs. (4)-(5)]
    "For each team, the fair gap is the difference in fair rewards which is equal to the fair reward (Equation 5) of the Fair algorithm minus the fair reward of the Efficient algorithm. A positive fair gap means that the Fair algorithm is fairer than the Efficient algorithm."

    The “fair reward” of the Fair algorithm is R(s)=1-F(s), which is exactly the objective that the policy is optimized to maximize (Eq. 5), while the Efficient algorithm maximizes E(s) (Eq. 4). Declaring a positive reward difference to be “fairer” is therefore a restatement of the optimization target, not an independent fairness criterion. Simulation Study #1 further filters teams using this fair gap, so the studied set is defined by the algorithm’s own reward.

  2. fitted input called prediction [Section 5.2.2 and Section 7.2.1, Eqs. (5), (10)-(11)]
    "The predicted fairness is the fair reward scaled to rating between 1 and 7. ... Figure 5 show the results for perceived fairness and predicted fairness from Simulation Study #2. The predicted fairness is the fair reward scaled to 1 to 7."

    For the Fair algorithm the fair reward is R(s)=1-F(s) (Eq. 5); for FEA, R(s)=λE(s)+(1-λ)(1-|FE(s)|') (Eq. 10), with the min/max scaling in Eq. (11) computed from user data (Section 6.2). Scaling this same reward to a 1–7 rating and calling it “predicted fairness” means the simulation’s prediction is the quantity the policy is built to maximize. A policy cannot confirm its own fairness premise by scoring high on its own objective; the only non-tautological evidence for perceived fairness is the participant ratings, which are small-n and untested statistically.

full rationale

The partial circularity is localized to the simulation-side fairness evaluations. The fair gap (Section 4.1) and the plotted “predicted fairness” (Sections 5.2.2 and 7.2.1) both reduce to the algorithms’ reward functions, so the simulation cannot independently validate the fair-equity metric or FEA. However, the central claim about human perception is not entirely forced by construction: User Study #2 collects external Likert ratings, which are genuine evidence, and the allocation metrics in Section 7.1.1 are objective behavioral outcomes. That external evidence is weak — n=7, no statistical tests (as Section 7.2.1 states), only the Mixed team type, and a 2-2 split on preference — and the paper itself lists “evaluating FEA within different team types” as future work, so the abstract’s “various human-agent team types” and “preferred” overstate the results. Those are scope and evidence-quality problems rather than circularity. No load-bearing self-citation chain is present: the cited prior metric [7] is used as a starting point, and the paper’s main test of it is human data, not the citation itself. Overall, one central simulation prediction reduces by construction, but the human-perception claim retains independent (if limited) content, giving a moderate score of 5 rather than 8–10.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The central claim rests on several hand-set values (lambda, fair-gap threshold, min-max normalization) and on importing equity theory without domain-specific validation. The capability and preference measurements are treated as ground truth, and only two of three team types are evaluated.

free parameters (4)
  • lambda (efficiency weight) = 0.70
    Hand-set in Eq. 10; no sensitivity analysis or justification for the value.
  • fair gap threshold = 0.60
    Teams filtered for fair gap > 0.60 in Section 5.1; chosen ad hoc to focus on teams where fairness matters.
  • min-max normalization constants = computed from User Study #1 data
    Eq. 11 uses min and max of |FE| from the Section 5.2 user sample, calibrating the reward on prior participant data.
  • capability rejection ranges = squares [0.20, 0.32], letters [0.63, 1.00]
    Simulation Study #2 rejection sampling is based on User Study #1 results (Section 7.1), restricting the team types explored.
assumptions (4)
  • domain assumption Task allocation can be modeled as an MDP with deterministic transitions and known rewards.
    Section 3 defines the MDP; assumes the robot has perfect knowledge of capabilities, preferences, and task progress.
  • domain assumption Capabilities and preferences are static and measurable before interaction.
    Section 4.2.1 measures them in a practice session; the model does not account for change, learning, or noise.
  • ad hoc to paper Adams' equity theory, where fairness is equality of outcome/input ratios, applies to human-agent task allocation.
    Section 6.1 imports the theory from sociology and fair ML without empirical validation in this domain.
  • domain assumption Human-agent team types follow a 50% Mixed, 25% Twins, 25% Negative prior.
    Section 6.3 assumes each teammate has a 50% chance of positive or negative capability-preference correlation; the prior is untested.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Fairness and Efficiency in Human-Agent Teams: An Iterative Algorithm Design Approach." pith.science (2026). https://pith.science/paper/6WJXHGUT

@misc{pith2026250516171,
  author       = {Pith},
  title        = {Pith review of: Fairness and Efficiency in Human-Agent Teams: An Iterative Algorithm Design Approach},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6WJXHGUT}},
  note         = {Machine review of arXiv:2505.16171}
}
read the original abstract

When agents interact with people as part of a team, fairness becomes an important factor. Prior work has proposed fairness metrics based on teammates' capabilities for task allocation within human-agent teams. However, most metrics only consider teammate capabilities from a third-person point of view (POV). In this work, we extend these metrics to include task preferences and consider a first-person POV. We leverage an iterative design method consisting of simulation data and human data to design a task allocation algorithm that balances task efficiency and fairness based on both capabilities and preferences. We first show that these metrics may not align with people's perceived fairness from a first-person POV. In light of this result, we propose a new fairness metric, fair-equity, and the Fair-Efficient Algorithm (FEA). Our findings suggest that an agent teammate who balances efficiency and fairness based on equity will be perceived to be fairer and preferred by human teammates in various human-agent team types. We suggest that the perception of fairness may also depend on a person's POV.

Figures

Figures reproduced from arXiv: 2505.16171 by the authors.

Figure 1
Figure 1. An example task allocation scenario that falls under our problem formulation. There are different task [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. User Study #1: An example of the collaborative session showing each teammate’s workspace and the [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Predicted fairness (scaled to rating between 1 and 7) from Simulation Study #1 and participants’ [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Simulation study #2 results for the Mixed team type. [PITH_FULL_IMAGE:figures/full_fig_p013_4.png]
Figure 5
Figure 5. Figure 5: Predicted fairness (scaled to rating between 1 and 7) from Simulation Study #2 and participants’ [PITH_FULL_IMAGE:figures/full_fig_p014_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 38 canonical work pages

  1. [1]

    J Stacy Adams. 1965. Inequity in social exchange. InAdvances in experimental social psychology. Vol. 2. Elsevier, 267–299

    work page 1965

  2. [2]

    Cristina Bicchieri. 1999. Local fairness.Philosophy and Phenomenological Research59, 1 (1999), 229–236

    work page 1999

  3. [3]

    Daniel Brugman, Christa Out, and John C Gibbs. 2016. Fairness and Trust in Developmental Psychology. InWomen and Children as Victims and Offenders: Background, Prevention, Reintegration. Springer, 265–289

    work page 2016

  4. [4]

    Simon Caton and Christian Haas. 2020. Fairness in Machine Learning: A Survey.arXiv preprint arXiv:2010.04053 (2020)

    arXiv 2020

  5. [5]

    Iadine Chadès, Guillaume Chapron, Marie-Josée Cros, Frédérick Garcia, and Régis Sabbadin. 2014. MDPtoolbox: a multi-platform toolbox to solve stochastic dynamic programming problems.Ecography37, 9 (2014), 916–920

    work page 2014

  6. [6]

    Mai Lee Chang, Zachary Pope, Elaine Schaertl Short, and Andrea Lockerd Thomaz. 2020. Defining Fairness in Human-Robot Teams. In2020 29th IEEE International Conference on Robot and Human Interactive Communication (RO-MAN). IEEE, 1251–1258

    work page 2020

  7. [7]

    Mai Lee Chang, Greg Trafton, J Malcolm McCurry, and Andrea Lockerd Thomaz. 2021. Unfair! Perceptions of Fairness in Human-Robot Teams. In2021 30th IEEE International Conference on Robot & Human Interactive Communication (RO-MAN). IEEE, 905–912

    work page 2021

  8. [8]

    Richard J Chen, Judy J Wang, Drew FK Williamson, Tiffany Y Chen, Jana Lipkova, Ming Y Lu, Sharifa Sahai, and Faisal Mahmood. 2023. Algorithmic fairness in artificial intelligence for medicine and healthcare.Nature Biomedical Engineering7, 6 (2023), 719–742

    work page 2023

Show all 43 references
  1. [9]

    Houston Claure, Mai Lee Chang, Seyun Kim, Daniel Omeiza, Martim Brandao, Min Kyung Lee, and Malte Jung

  2. [10]

    Houston Claure, Yifang Chen, Jignesh Modi, Malte Jung, and Stefanos Nikolaidis. 2020. Multi-Armed Bandits with Fairness Constraints for Distributing Resources to Human Teammates. InProceedings of the 2020 ACM/IEEE International Conference on Human-Robot Interaction. 299–308

    2020

  3. [11]

    Houston Claure, Seyun Kim, René F Kizilcec, and Malte Jung. 2023. The social consequences of machine allocation behavior: Fairness, interpersonal perceptions and performance.Computers in Human Behavior146 (2023), 107628

    2023

  4. [12]

    Sam Corbett-Davies, Emma Pierson, Avi Feller, Sharad Goel, and Aziz Huq. 2017. Algorithmic decision making and the cost of fairness. InProceedings of the 23rd acm sigkdd international conference on knowledge discovery and data mining. 797–806

    2017

  5. [13]

    David De Cremer and Tom R Tyler. 2007. The effects of trust in authority and procedural fairness on cooperation. Journal of applied psychology92, 3 (2007), 639. Fairness and Efficiency in Human-Agent Teams 17

    2007

  6. [14]

    Wesley Hanwen Deng, Manish Nagireddy, Michelle Seng Ah Lee, Jatinder Singh, Zhiwei Steven Wu, Kenneth Holstein, and Haiyi Zhu. 2022. Exploring how machine learning practitioners (try to) use fairness toolkits. InProceedings of the 2022 ACM Conference on Fairness, Accountabilit...

    2022

  7. [15]

    Sanghamitra Dutta, Dennis Wei, Hazar Yueksel, Pin-Yu Chen, Sijia Liu, and Kush Varshney. 2020. Is there a trade-off between fairness and accuracy? A perspective using mismatched hypothesis testing. InInternational Conference on Machine Learning. PMLR, 2803–2813

    2020

  8. [16]

    Cynthia Dwork, Moritz Hardt, Toniann Pitassi, Omer Reingold, and Richard Zemel. 2012. Fairness through awareness. InProceedings of the 3rd innovations in theoretical computer science conference. 214–226

    2012

  9. [17]

    Jodi Forlizzi. 2018. Moving beyond user-centered design.interactions25, 5 (2018), 22–23

    2018

  10. [18]

    Adam D Galinsky and Gordon B Moskowitz. 2000. Perspective-taking: decreasing stereotype expression, stereotype accessibility, and in-group favoritism.Journal of personality and social psychology78, 4 (2000), 708

    2000

  11. [19]

    Matthew Gombolay, Anna Bair, Cindy Huang, and Julie Shah. 2017. Computational design of mixed-initiative human– robot teaming that considers human factors: situational awareness, workload, and workflow preferences.International Journal of Robotics Research36, 5-7 (2017), 597–617

    2017

  12. [20]

    Matthew C Gombolay, Reymundo A Gutierrez, Shanelle G Clarke, Giancarlo F Sturla, and Julie A Shah. 2015. Decision- making authority, team efficiency and human worker satisfaction in mixed human–robot teams.Autonomous Robots 39, 3 (2015), 293–312

    2015

  13. [21]

    Matthew Craig Gombolay, Cindy Huang, and Julie Shah. 2015. Coordination of human-robot teaming with human task preferences. In2015 AAAI Fall Symposium Series

    2015

  14. [22]

    Paul S Goodman and Abraham Friedman. 1971. An examination of Adams’ theory of inequity.Administrative Science Quarterly(1971), 271–288

    1971

  15. [23]

    Soheil Habibian and Dylan P Losey. 2022. Encouraging human interaction with robot teams: Legible and fair subtask allocations.IEEE Robotics and Automation Letters7, 3 (2022), 6685–6692

    2022

  16. [24]

    Moritz Hardt, Eric Price, and Nati Srebro. 2016. Equality of Opportunity in Supervised Learning. InNIPS

    2016

  17. [25]

    Ben Hutchinson and Margaret Mitchell. 2019. 50 years of test (un) fairness: Lessons for machine learning. InProceedings of the Conference on Fairness, Accountability, and Transparency. 49–58

    2019

  18. [26]

    Malte F Jung, Dominic DiFranzo, Solace Shen, Brett Stoll, Houston Claure, and Austin Lawrence. 2020. Robot-Assisted Tower Construction—A Method to Study the Impact of a Robot’s Allocation Behavior on Interpersonal Dynamics and Collaboration in Groups.ACM Transactions on Human-...

    2020

  19. [27]

    Joon Sik Kim, Jiahao Chen, and Ameet Talwalkar. 2020. Fact: A diagnostic for group fairness trade-offs. InInternational Conference on Machine Learning. PMLR, 5264–5274

    2020

  20. [28]

    Irving M Lane and Lawrence A Messe. 1971. Equity and the distribution of rewards.Journal of Personality and Social Psychology20, 1 (1971), 1

    1971

  21. [29]

    Tai Le Quy, Arjun Roy, Vasileios Iosifidis, Wenbin Zhang, and Eirini Ntoutsi. 2022. A survey on datasets for fairness- aware machine learning.Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery12, 3 (2022), e1452

    2022

  22. [30]

    Aditya Krishna Menon and Robert C Williamson. 2018. The cost of fairness in binary classification. InConference on Fairness, Accountability and Transparency. 107–118

    2018

  23. [31]

    Manisha Natarajan, Esmaeil Seraj, Batuhan Altundas, Rohan Paleja, Sean Ye, Letian Chen, Reed Jensen, Kimber- lee Chestnut Chang, and Matthew Gombolay. 2023. Human-Robot Teaming: Grand Challenges.Current Robotics Reports(2023), 1–20

    2023

  24. [32]

    Stefanos Nikolaidis and Julie Shah. 2013. Human-robot cross-training: computational formulation, modeling and evaluation of a human team training strategy. InProceedings of the 8th ACM/IEEE International Conference on Human- Robot Interaction. IEEE Press, 33–40

    2013

  25. [33]

    Luca Oneto and Silvia Chiappa. 2020. Fairness in Machine Learning. InRecent Trends in Learning From Data. Springer, 155–196

    2020

  26. [34]

    Tiago P Pagano, Rafael B Loureiro, Fernanda VN Lisboa, Rodrigo M Peixoto, Guilherme AS Guimarães, Gustavo OR Cruz, Maira M Araujo, Lucas L Santos, Marco AS Cruz, Ewerton LS Oliveira, et al. 2023. Bias and unfairness in machine learning models: a systematic review on datasets, ...

    2023

  27. [35]

    Dana Pessach and Erez Shmueli. 2022. A review on fairness in machine learning.ACM Computing Surveys (CSUR)55, 3 (2022), 1–44

    2022

  28. [36]

    Dominik Riedelbauch, Nico Höllerich, and Dominik Henrich. 2023. Benchmarking Teamwork of Humans and Cobots– An Overview of Metrics, Strategies, and Tasks.IEEE Access(2023)

    2023

  29. [37]

    1966.Relative deprivation and social justice: A study of attitudes to social inequality in twentieth-century England

    Walter Garrison Runciman and Baron Runciman. 1966.Relative deprivation and social justice: A study of attitudes to social inequality in twentieth-century England. Vol. 13. University of California Press Berkeley

    1966

  30. [38]

    Sarah Sebo, Brett Stoll, Brian Scassellati, and Malte F Jung. 2020. Robots in Groups and Teams: A Literature Review. Proceedings of the ACM on Human-Computer Interaction4, CSCW2 (2020), 1–36. 18 Mai Lee Chang et al

    2020

  31. [39]

    Mingyang Wan, Daochen Zha, Ninghao Liu, and Na Zou. 2023. In-processing modeling techniques for machine learning fairness: A survey.ACM Transactions on Knowledge Discovery from Data17, 3 (2023), 1–27

    2023

  32. [40]

    Franziska Doris Wolf and Ruth Stock-Homburg. 2020. Human-Robot Teams: A Review. InInternational Conference on Social Robotics. Springer, 246–258

    2020

  33. [41]

    Qian Yang, Aaron Steinfeld, Carolyn Rosé, and John Zimmerman. 2020. Re-examining whether, why, and how human- AI interaction is uniquely difficult to design. InProceedings of the 2020 chi conference on human factors in computing systems. 1–13

    2020

  34. [42]

    Han Zhao and Geoff Gordon. 2019. Inherent tradeoffs in learning fair representations. InAdvances in neural information processing systems. 15675–15685

    2019

  35. [2022]

    In2022 17th ACM/IEEE International Conference on Human-Robot Interaction (HRI)

    Fairness and Transparency in Human-Robot Interaction. In2022 17th ACM/IEEE International Conference on Human-Robot Interaction (HRI). IEEE, 1244–1246

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.