Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

Simulating Errors in Touchscreen Typing

T0 review · 4 major / 5 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read This paper introduces Typoist, the first typing model that simulates slips, lapses, and mistakes in touchscreen typing and reproduces human error distributions across diverse user groups.

desk verdict A useful modeling framework and benchmark, but the evaluation loop is circular: error parameters are fitted on the same metrics used to claim the model reproduces human errors. read the letter →

arxiv 2502.03560 v1 pith:ZHWNX5D6 submitted 2025-02-05 cs.HC

Danqing Shi , Yujun Zhu , Francisco Erivaldo Fernandes Junior , Shumin Zhai , Antti Oulasvirta This is my paper · ORCID
classification cs.HC
keywords touchscreentypinghumanerrorsusersimulationsupervisorycontrolcomputationalrationalityPOMDPslipslapsesmistakeserrorbenchmark
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 introduces Typoist, a computational model that takes a phrase, a keyboard design, and user parameters and simulates how a person would type it on a touchscreen, including where their eyes and fingers move and which errors they make. The paper's central claim is that Typoist is the first typing model to cover all three cognitive error sources — slips, lapses, and mistakes — not just motor slips, and to produce insertion, omission, substitution, and transposition errors as a result. Across a new benchmark with data from young adults, elderly users, and people with Parkinson's disease, Typoist reproduces error rates and correction behavior within one standard deviation of human data on most metrics, while the prior CRTypist model reproduces only a subset. The practical payoff is that keyboard and text-entry designs could be evaluated in simulation before running user studies, and individual differences could be tuned through a small set of interpretable parameters.

What carries the argument

The central object is the supervisory control loop, formulated as a POMDP in which a reinforcement-learned controller sets goals for a vision module, a finger module, and working memory while the simulated cognitive environment injects errors. The error-generating mechanisms are parameterized by a small set of functions: the WHo model for finger endpoint spread; forgetting with probability $P(t)=1-e^{-kt}$, where $t$ is time since the last proofreading; proofreading miss probability $P_{\text{obs--text}}=p_0 e^{-T}$, where longer proofreading raises accuracy; and a constant finger-observation miss probability $P_{\text{obs--finger}}$. These parameters scale the user's capabilities, and a two-loop Bayesian optimization fits both training hyperparameters and user-group parameters so that the same architecture covers different populations. The reward $R=(1-\text{Err}^\alpha)-w t$ encodes the speed–accuracy tradeoff that drives the controller's allocation of attention.

What would settle it

A study that measured forgetting rates while holding time since the last proofreading constant — for example, varying phrase difficulty or cognitive load — and found different lapse rates would falsify the exponential forgetting assumption; similarly, a proofreading experiment in which miss rates do not follow $p_0 e^{-T}$ as reading time varies would falsify the mistakes mechanism.

Watch

Extended reading notes

Core claim

Typoist models typing as a partially observable Markov decision process under a supervisory controller that decides where to look and what to type, while the internal environment of the model contains noisy cognitive capabilities that generate errors. Motor execution noise produces slips such as hitting a neighboring key, unintentional double taps, and swapped keypress order; a forgetting probability that grows with time since the last proofreading produces lapses; and imperfect proofreading plus imperfect observation of the finger during gaze guidance produce mistakes. The controller optimizes a reward that balances error rate against time, so the model chooses speed–accuracy tradeoffs the way a user might. The paper claims that with parameter values fitted to different user groups, Typoist generates error distributions and correction strategies comparable to those of young adults, elderly typists, and individuals with Parkinson's, and that it captures the effects of autocorrection better than the earlier CRTypist model.

Load-bearing premise

The load-bearing premise is that the error mechanisms can be captured by simple formulas — forgetting grows only with time since last proofreading, proofreading accuracy grows exponentially with reading time, and finger-location misses are a constant — so if those formulas are wrong, the model's fit to observed error rates would not show it simulates the true causes.

Editorial extensions

If this is right

  • Keyboard designs could be evaluated for fault tolerance in simulation before a user study, including error types that prior models could not generate.
  • Practitioners could generate synthetic typing error data for groups that are hard to recruit, such as users with Parkinson's or elderly users, by setting the model's cognitive parameters.
  • Error-handling behavior, such as immediate versus delayed backspacing, can be predicted together with typing speed, so design changes can be judged on correction effort rather than raw error counts alone.
  • Because every simulated typographical error is traced to a cognitive mechanism, the model offers a glass-box alternative to purely statistical text-error generators.

Reading between the lines

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

  • A testable extension is to check whether the simple functional forms transfer to new phrase sets and keyboards; if forgetting depends on more than time since the last proofreading, the parameter fits may not generalize.
  • The model's rule-based autocorrection is likely too simple for real-world errors such as space-key confusion, where the decoder itself creates the error; simulating the decoder's behavior would be a natural next step.
  • The resource-rationality story implies that a given user's speed–accuracy preference should be stable across keyboard layouts, which could be tested with the same benchmark data.
  • The same supervisory-control architecture could be applied to other error-prone input tasks, such as gesture typing or pointing, where slips, lapses, and mistakes also occur.
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

4 major / 5 minor

Summary. Typoist extends the CRTypist computational-rationality model of touchscreen typing by adding mechanisms for slips, lapses, and mistakes, and by formulating typing as a POMDP in which a supervisory controller allocates gaze and finger resources under noisy perception, motor control, and memory. The paper introduces a three-level benchmark (no correction, manual correction, autocorrection) with datasets from young adults, elderly users, Parkinson's users, Finnish typists, and Gboard typists, and reports that Typoist reproduces human error rates, error corrections, and speed-accuracy trade-offs better than CRTypist. The authors also provide an interactive visualization tool and release the model.

Significance. If the model's mechanisms are validated, this would be a substantial advance: it is the first typing simulation to move beyond motor slips and to cover the three classical cognitive error categories in a unified, computationally rational framework. The paper also contributes a public benchmark organized by error-correction condition, a useful artifact for future text-entry research, and a reproducible RL-plus-Bayesian-optimization pipeline. The strongest parts are the breadth of the model architecture and the honest reporting of several residual mismatches. However, the central empirical claim—that agreement with human data validates the proposed cognitive mechanisms—is currently weakened by an in-sample parameter-fitting evaluation, so the significance is contingent on additional held-out or otherwise non-circular validation.

major comments (4)
  1. [§3.3 and §5.2] The evaluation is in-sample with respect to the fitted parameters. Section 3.3 describes an inner-loop optimization that adapts user-group parameters, and Section 5.2 states that optimization minimizes differences in typing speed, error rates, and backspacing. Table 1 then reports those same metrics as evidence that Typoist 'closely reproduce[s] human-like behavior.' Because the parameters are fit to the very metrics used for evaluation, the agreement in Table 1 does not yet discriminate mechanistic simulation from flexible curve fitting. I would like to see a held-out evaluation: fit parameters on one set of phrases or participants and predict a disjoint set, or predict a new keyboard or autocorrection condition without refitting, plus a parameter-identifiability or sensitivity analysis showing that the fitted values are constrained by the data rather than merely absorbing model misspecification.
  2. [Table 1, Level 0 and Level 1 rows] Substantial deviations remain even on the fitted metrics, and these undermine the blanket claim in Section 5 that the model 'can closely reproduce human-like behavior at each level.' Specifically, Parkinson's insertion errors are simulated at 2.75% versus 6.63% human (M=6.63, SD=1.24), elderly insertion errors at 0.48% versus 4.60% human, and Gboard substitution errors at 6.38% versus 1.78% human. Sections 5.1 and 5.2 acknowledge these exceptions, but the abstract and Section 6 still present the fit as close. These mismatches should be reported as failures and analyzed structurally (e.g., is the double-tap probability P(v)=1-e^{-k*v} unable to reach the observed Parkinson's insertion rate?), rather than listed as minor residuals.
  3. [§5.1, Elderly users] The comparison for elderly users relies on an unjustified assumption: the text says the reference data do not include standard deviations and that the authors 'assume the SD to be 1%' in order to declare that three error rates fall within range. Since 1% is arbitrarily chosen and differs from the SDs reported for other groups, the claim that the model matches elderly error rates is not supported. Either report the actual variability or refrain from making statistical-fit claims for this group.
  4. [§5.2, Bayes factor analysis] The Bayes factor comparison between Typoist and CRTypist is presented only as aggregate counts (e.g., '6/7 show support for H0'), without listing the Bayes factors for most metrics or the prior used in the Bayesian t-test. Given that the same data were used for parameter fitting, these tests do not address the central circularity concern; additionally, the Gboard dataset comparison appears to use only 30 simulated data points against 5,140 human trajectories, which makes the test underpowered. Please clarify the sample sizes and report the full set of Bayes factors and priors.
minor comments (5)
  1. [Table 1 header] The table header contains a garbled fragment ('/edtmanual error correction is allowed; gaze data is included;') that should be cleaned up, and the superscript markers used to indicate conditions are not defined consistently.
  2. [§3.1.1, equation] The WHo model equation is typeset incorrectly: '(y-y0)1-kα(x-x0)kα = FK' is missing the operators and parentheses needed to be readable. Please use the standard mathematical notation from Guiard and Rioul.
  3. [References] References [45] and [46] are duplicates of the same PPO arXiv paper; one should be removed or differentiated.
  4. [§6.1] The Discussion introduces the term 'commission' errors, but the paper elsewhere consistently uses 'insertion' errors; please align the terminology.
  5. [§5.3] The sentence 'the model exhibited a slight increase in typing speed when autocorrection is not enabled' is ambiguous because the comparison is between the Level 2 model and the Level 1 model, not a within-condition manipulation; please rephrase for clarity.

Circularity Check

1 steps flagged · score 6.0 of 10

Reported error reproduction is a training-set fit: per-user-group parameters are optimized on the same benchmark metrics later presented as predicted matches.

  1. fitted input called prediction [Section 5.2 (Level 1); see also Section 5.1, Section 3.3, and Table 1]
    "optimization of all human parameters for the model for the target user group was handled by minimizing the differences in typing speed, error rates, and the amount of backspacing."

    The human parameters that control the error mechanisms (forgetting rate k, proofreading miss p0, finger-miss probability P_obs_finger, motor noise FK, and reward weights) are optimized per user group to minimize the difference between simulated and human behavior on the very metrics reported in Table 1. Section 5.1 similarly says, 'After adjusting cognitive parameters for these groups, we ran simulations with the same number of independent episodes for each group, then compared the results.' The close matches in Table 1 are therefore consequences of fitting, not independent predictions, and they cannot validate the claimed lapse/mistake mechanisms without held-out phrases, participants, keyboard layouts, or a parameter-identifiability analysis.

full rationale

The central quantitative evidence in Table 1 is not an independent test of Typoist's cognitive mechanisms: the same benchmark metrics (typing speed, error rates, backspacing) are used both as the optimization objective for per-group human parameters and as the outcome measures reported as successful reproduction. This is the core circular step. The paper's Bayes-factor analysis in Section 5.2 compares the fitted simulations to the human data after parameter optimization, which again treats a training-set fit as predictive evidence. No held-out phrases, held-out participants, novel keyboard condition, or parameter-identifiability check is reported, so the match cannot distinguish mechanistic simulation from flexible curve fitting. The model is not wholly definitional: the architecture integrates slips, lapses, and mistakes, and the trajectory-level and autocorrection behavior have independent content. However, the load-bearing claim that Typoist 'reproduces' human error distributions is partly forced by the fitting procedure. Self-citations to CRTypist are present and the model is built on CRTypist's internal environment, but the decisive circularity here is the fitted-input-called-prediction loop, not the self-citation chain itself.

Assumptions & free parameters 7 free parameters · 6 assumptions · 0 invented entities

The model's predictions depend on several free parameters fitted to the benchmark data, on background assumptions about error taxonomy, motor control, computational rationality, and on the specific POMDP formulation. No new physical or conceptual entities are introduced; the invented_entities list is empty.

free parameters (7)
  • FK (finger capability) = Not reported in text; optimized per user group
    Controls motor control noise in the Weighted Homographic model, causing substitution and omission errors.
  • k (double tap and command swap) = Not reported in text; optimized per user group
    Rate parameter in P(v)=1-e^{-k*v} controlling unintentional insertions and transpositions.
  • k (lapse forgetting) = Not reported in text; optimized per user group
    Rate constant in P(t)=1-e^{-k*t} for forgetting to send a motor command, leading to omissions.
  • p0 (proofreading miss baseline) = Not reported in text; optimized per user group
    Base probability in P_obs_text=p0*e^{-T} of missing a typo during proofreading.
  • P_obs_finger = Not reported in text; optimized per user group
    Constant probability of missing a finger slip during visual guidance.
  • alpha (error sensitivity) = Not reported in text; optimized per user group
    Exponent in reward function R=(1-Err^alpha)-w*t controlling sensitivity to errors.
  • w (time weight) = Not reported in text; optimized per user group
    Weight on time in the reward function, balancing speed and accuracy.
assumptions (6)
  • domain assumption Reason's taxonomy of human error: slips, lapses, mistakes.
    Used in Section 1 and 2 to categorize error mechanisms and map them to typographical errors.
  • domain assumption Information-processing view of human error (Wickens et al.) maps error types to stages: interpretation, intention, execution.
    Adopted in Section 3.1 to justify the mapping from mechanisms to slips, lapses, and mistakes.
  • standard math The Weighted Homographic (WHo) model describes finger motor noise.
    Used in Section 3.1.1 to model substitution and omission errors from motor slips; taken from prior literature.
  • domain assumption Human behavior can be modeled as maximizing expected utility under cognitive bounds (computational rationality).
    Foundation of the POMDP formulation in Section 3.3; prior work in the same group.
  • ad hoc to paper Typing can be formulated as a POMDP with pixel-level state, belief observations, finger/gaze action spaces, and a reward trading off errors and time.
    The specific POMDP definition in Section 3.2 is a modeling assumption not directly derived from data or theory.
  • domain assumption The datasets used in the TypingError benchmark accurately represent the human error distributions they are claimed to measure.
    The benchmark in Section 4 relies on four external datasets; their correctness is assumed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Simulating Errors in Touchscreen Typing." pith.science (2026). https://pith.science/paper/ZHWNX5D6

@misc{pith2026250203560,
  author       = {Pith},
  title        = {Pith review of: Simulating Errors in Touchscreen Typing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZHWNX5D6}},
  note         = {Machine review of arXiv:2502.03560}
}
read the original abstract

Empirical evidence shows that typing on touchscreen devices is prone to errors and that correcting them poses a major detriment to users' performance. Design of text entry systems that better serve users, across their broad capability range, necessitates understanding the cognitive mechanisms that underpin these errors. However, prior models of typing cover only motor slips. The paper reports on extending the scope of computational modeling of typing to cover the cognitive mechanisms behind the three main types of error: slips (inaccurate execution), lapses (forgetting), and mistakes (incorrect knowledge). Given a phrase, a keyboard, and user parameters, Typoist simulates eye and finger movements while making human-like insertion, omission, substitution, and transposition errors. Its main technical contribution is the formulation of a supervisory control problem wherein the controller allocates cognitive resources to detect and fix errors generated by the various mechanisms. The model generates predictions of typing performance that can inform design, for better text entry systems.

Figures

Figures reproduced from arXiv: 2502.03560 by the authors.

Figure 1
Figure 1. We introduce the first model covering a wide spectrum of errors known to be commonplace in typing, [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. a) An information-processing view of human er [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. A simulation example involving multiple mechanisms that generate various text errors and corrections. In typing [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Visualization tool for exploring simulations. a) Via [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Typing speed vs. error corrections. The figure shows the speed–accuracy tradeoff in both human data and the [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Synthesizing and Adapting Error Correction Data for Mobile Large Language Model Applications

    cs.LG 2025-05 conditional novelty 5.0 of 10

    A production mobile-keyboard error-correction system that synthesizes large LLM-generated training data, reweights it with a differentially private federated small LM, and fine-tunes a billion-parameter LLM via LoRA r...

Reference graph

Works this paper leans on

60 extracted references · 55 canonical work pages · cited by 1 Pith paper

  1. [1]

    Anderson, Daniel Bothell, Michael D

    John R. Anderson, Daniel Bothell, Michael D. Byrne, Scott Douglass, Christian Lebiere, and Yulin Qin. 2004. An integrated theory of the mind. Psychological Review 111, 4 (2004), 1036–1060

    work page 2004

  2. [2]

    Marcin Andrychowicz, Anton Raichuk, Piotr Stańczyk, Manu Orsini, Sertan Girgin, Raphael Marinier, Léonard Hussenot, Matthieu Geist, Olivier Pietquin, Marcin Michalski, Sylvain Gelly, and Olivier Bachem. 2020. What Matters In On-Policy Reinforcement Learning? A Large-Scale Empirical Study. http: //arxiv.org/abs/2006.05990

    arXiv 2020

  3. [3]

    Ahmed Sabbir Arif, Sunjun Kim, Wolfgang Stuerzlinger, Geehyuk Lee, and Ali Mazalek. 2016. Evaluation of a smart-restorable backspace technique to facilitate text entry error correction. In Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems . 5151–5162

    work page 2016

  4. [4]

    Ahmed Sabbir Arif and Wolfgang Stuerzlinger. 2009. Analysis of text entry performance metrics. In Proceedings of the IEEE Toronto International Conference – Science and Technology for Humanity (TIC-STH 2009) . IEEE, New York, NY, 100–105

    work page 2009

  5. [5]

    Nikola Banovic, Tovi Grossman, and George Fitzmaurice. 2013. The effect of time- based cost of error in target-directed pointing tasks. In CHI ’13: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems . ACM, 1373–1382

    work page 2013

  6. [6]

    Dey, and Jennifer Mankoff

    Nikola Banovic, Varun Rao, Abinaya Saravanan, Anind K. Dey, and Jennifer Mankoff. 2017. Quantifying aversion to costly typing errors in expert mobile text entry. In CHI ’17: Proceedings of the 2017 CHI Conference on Human Factors in Computing Systems. ACM, 4229–4241

    work page 2017

  7. [7]

    Dey, and Jennifer Mankoff

    Nikola Banovic, Ticha Sethapakdi, Yasasvi Hari, Anind K. Dey, and Jennifer Mankoff. 2019. The limits of expert text entry speed on mobile keyboards with autocorrect. In MobileHCI ’19: Proceedings of the 21st International Conference on Human–Computer Interaction with Mobile Devices and Services . ACM, Article 15

    work page 2019

  8. [8]

    Patrick Baudisch and Gerry Chu. 2009. Back-of-device interaction allows creating very small touch devices. In CHI ’09: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems . 1923–1932

    work page 2009

Show all 60 references
  1. [9]

    Xiaojun Bi, Yang Li, and Shumin Zhai. 2013. FFitts law: Modeling finger touch with Fitts’ law. InCHI ’13: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems. ACM, New York, NY, 1363–1372

    2013

  2. [10]

    Xiaojun Bi and Shumin Zhai. 2013. Bayesian Touch: A statistical criterion of target selection with finger touch. In Proceedings of the 26th Annual ACM Symposium on User Interface Software and Technology . 51–60

    2013

  3. [11]

    Matthew Michael Botvinick. 2012. Hierarchical reinforcement learning and decision making. Current Opinion in Neurobiology 22, 6 (2012), 956–962

    2012

  4. [12]

    Tania Cerni, Marieke Longcamp, and Remo Job. 2016. Two thumbs and one index: A comparison of manual coordination in touch-typing and mobile-typing. Acta Psychologica 167 (June 2016), 16–23

    2016

  5. [13]

    Suyog Chandramouli, Danqing Shi, Aini Putkonen, Sebastiaan De Peuter, Shan- shan Zhang, Jussi Jokinen, Andrew Howes, and Antti Oulasvirta. 2024. A work- flow for building computationally rational models of human behavior. Computa- tional Brain & Behavior 7 (2024), 399–419

    2024

  6. [14]

    Vivek Dhakal, Anna Maria Feit, Per Ola Kristensson, and Antti Oulasvirta. 2018. Observations on typing from 136 million keystrokes. In CHI ’18: Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems . Article 646

    2018

  7. [15]

    Donaldson, Janet M

    Molla S. Donaldson, Janet M. Corrigan, and Linda T. Kohn. 2000. To err is human: Building a safer health system. (2000)

    2000

  8. [16]

    Logan Engstrom, Andrew Ilyas, Shibani Santurkar, Dimitris Tsipras, Firdaus Janoos, Larry Rudolph, and Aleksander Madry. 2020. Implementation matters in deep RL: A case study on PPO and TRPO. In International Conference on Learning Representations. https://openreview.net/forum?...

    2020

  9. [17]

    Paul M. Fitts. 1954. The information capacity of the human motor system in controlling the amplitude of movement. Journal of Experimental Psychology 47, 6 (1954), 381–391

    1954

  10. [18]

    Jerry A. Fodor. 1983. The modularity of mind . MIT Press. Simulating Errors in Touchscreen Typing CHI ’25, April 26-May 1, 2025, Yokohama, Japan

    1983

  11. [19]

    Frank and David Badre

    Michael J. Frank and David Badre. 2012. Mechanisms of hierarchical reinforce- ment learning in corticostriatal circuits 1: Computational analysis. Cerebral Cortex 22, 3 (2012), 509–526

    2012

  12. [20]

    Peter I. Frazier. 2018. Bayesian optimization. In Recent advances in optimization and modeling of contemporary problems , Esma Gel, Lewis Ntaimo, Douglas Shier, and Harvey J. Greenberg (Eds.). INFORMS, 255–278. https://doi.org/10.1287/ educ.2018.0188

    2018

  13. [21]

    Gentner, Jonathan T

    Donald R. Gentner, Jonathan T. Grudin, Serge Larochelle, Donald A. Norman, and David E. Rumelhart. 1983. A glossary of terms including a classification of typing errors. In Cognitive aspects of skilled typewriting , William H. Cooper (Ed.). Springer, 39–43

    1983

  14. [22]

    Yves Guiard and Olivier Rioul. 2015. A mathematical description of the speed/accuracy trade-off of aimed movement. In British HCI ’15: Proceedings of the 2015 British HCI Conference . 91–100

    2015

  15. [23]

    Haber and Robert M

    Ralph N. Haber and Robert M. Schindler. 1981. Error in proofreading: Evidence of syntactic control of letter processing? Journal of Experimental Psychology: Human Perception and Performance 7, 3 (1981), 573–579

    1981

  16. [24]

    Brewster, and Jody Johnston

    Eve Hoggan, Stephen A. Brewster, and Jody Johnston. 2008. Investigating the effectiveness of tactile feedback for mobile touchscreens. InCHI ’08: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems. ACM, 1573–1582

    2008

  17. [25]

    Christian Holz and Patrick Baudisch. 2010. The generalized perceived input point model and how to double touch accuracy by extracting fingerprints. In CHI ’10: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems . 581–590

    2010

  18. [26]

    Christian Holz and Patrick Baudisch. 2011. Understanding touch. In CHI ’11: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems . 2501–2510

    2011

  19. [27]

    Xinhui Jiang, Yang Li, Jussi P. P. Jokinen, Viet Ba Hirvola, Antti Oulasvirta, and Xiangshi Ren. 2020. How we type: Eye and finger movement strategies in mobile typing. In CHI ’20: Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems. 1–14

    2020

  20. [28]

    Jussi Jokinen, Aditya Acharya, Mohammad Uzair, Xinhui Jiang, and Antti Oulasvirta. 2021. Touchscreen typing as optimal supervisory control. In CHI ’21: Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems . ACM, Article 720

    2021

  21. [29]

    Jussi Jokinen, Sayan Sarcar, Antti Oulasvirta, Chaklam Silpasuwanchai, Zhenxin Wang, and Xiangshi Ren. 2017. Modelling learning of new keyboard layouts. In CHI ’17: Proceedings of the 2017 CHI Conference on Human Factors in Computing Systems (Denver, CO). ACM, New York, NY, 4203–4215

    2017

  22. [30]

    Griffiths

    Falk Lieder and Thomas L. Griffiths. 2020. Resource-rational analysis: Under- standing human cognition as the optimal use of limited computational resources. Behavioral and brain sciences 43, Article e1 (2020)

    2020

  23. [31]

    Logan and Matthew J

    Gordon D. Logan and Matthew J. C. Crump. 2010. Cognitive illusions of author- ship reveal hierarchical error detection in skilled typists. Science 330, 6004 (2010), 683–686

    2010

  24. [32]

    Scott MacKenzie and R

    I. Scott MacKenzie and R. William Soukoreff. 2002. Text entry for mobile com- puting: Models and methods, theory and practice. Human–Computer Interaction 17, 2–3 (2002), 147–198

    2002

  25. [33]

    Scott MacKenzie and R

    I. Scott MacKenzie and R. William Soukoreff. 2003. Phrase sets for evaluating text entry techniques. In CHI ’03 Extended Abstracts on Human Factors in Computing Systems. 754–755

    2003

  26. [34]

    Hugo Nicolau and Joaquim Jorge. 2012. Elderly text-entry performance on touchscreens. In ASSETS ’12: Proceedings of the 14th International ACM SIGACCESS Conference on Computers and Accessibility . ACM, 127–134

    2012

  27. [35]

    Donald A. Norman. 1988. The psychology of everyday things . Basic Books

    1988

  28. [36]

    Antti Oulasvirta, Jussi P. P. Jokinen, and Andrew Howes. 2022. Computational rationality as a theory of interaction. In CHI ’22: Proceedings of the 2022 CHI Conference on Human Factors in Computing Systems . ACM, Article 359

    2022

  29. [37]

    Tom Le Paine, Cosmin Paduraru, Andrea Michi, Caglar Gulcehre, Konrad Zolna, Alexander Novikov, Ziyu Wang, and Nando de Freitas. 2020. Hyperparameter selection for offline reinforcement learning. http://arxiv.org/abs/2007.09055

    2020 arXiv

  30. [38]

    Kseniia Palin, Anna Maria Feit, Sunjun Kim, Per Ola Kristensson, and Antti Oulasvirta. 2019. How do people type on mobile devices? Observations from a study with 37,000 volunteers. In Proceedings of the 21st International Conference on Human–Computer Interaction with Mobile De...

    2019

  31. [39]

    Maura Pilotti, Martin Chodorow, and Frances Schauss. 2009. Text familiarity, word frequency, and sentential constraints in error detection. Perceptual and Motor Skills 109, 3 (2009), 627–645

    2009

  32. [40]

    Svetlana Pinet and Nazbanou Nozari. 2022. Correction without consciousness in complex tasks: Evidence from typing. Journal of Cognition 5, 1, Article 11 (2022)

    2022

  33. [41]

    James Reason. 1990. Human error. Cambridge University Press

    1990

  34. [42]

    Rouder, Paul L

    Jeffrey N. Rouder, Paul L. Speckman, Dongchu Sun, Richard D. Morey, and Geof- frey Iverson. 2009. Bayesian t tests for accepting and rejecting the null hypothesis. Psychonomic Bulletin & Review 16 (2009), 225–237

    2009

  35. [43]

    Quentin Roy, Sébastien Berlioux, Géry Casiez, and Daniel Vogel. 2021. Typing efficiency and suggestion accuracy influence the benefits and adoption of word suggestions. In CHI ’21: Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems. ACM, Article 714

    2021

  36. [44]

    Rumelhart and Donald A

    David E. Rumelhart and Donald A. Norman. 1982. Simulating a skilled typist: A study of skilled cognitive-motor performance. Cognitive Science 6, 1 (1982)

    1982

  37. [46]

    John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov

  38. [47]

    Danqing Shi, Yujun Zhu, Jussi Jokinen, Aditya Acharya, Aini Putkonen, Shumin Zhai, and Antti Oulasvirta. 2024. CRTypist: Simulating touchscreen typing behavior via computational rationality. In Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems

    2024

  39. [48]

    http://arxiv.org/abs/1707.06347 arXiv:1707.06347 [cs]

    Proximal policy optimization algorithms. http://arxiv.org/abs/1707.06347 arXiv:1707.06347 [cs]

    arXiv

  40. [49]

    Yuntao Wang, Ao Yu, Xin Yi, Yuanwei Zhang, Ishan Chatterjee, Shwetak Patel, and Yuanchun Shi. 2021. Facilitating text entry on smartphones with QWERTY keyboard for users with Parkinson’s disease. In CHI ’21: Proceedings of the 2021 CHI Conference on Human Factors in Computing ...

    2021

  41. [50]

    Siek, Yvonne Rogers, and Kay H

    Katie A. Siek, Yvonne Rogers, and Kay H. Connelly. 2005. Fat finger worries: How older and younger users physically interact with PDAs. In Human–computer interaction – INTERACT 2005: IFIP TC13 International Conference, Rome, Italy, September 12–16, 2005. Proceedings 10 . Sprin...

    2005

  42. [51]

    Wickens, William S

    Christopher D. Wickens, William S. Helton, Justin G. Hollands, and Simon Ban- bury. 2021. Engineering psychology and human performance . Routledge

    2021

  43. [52]

    Daryl Weir, Henning Pohl, Simon Rogers, Keith Vertanen, and Per Ola Kristensson

  44. [53]

    Wobbrock, Edward Cutrell, Susumu Harada, and I

    Jacob O. Wobbrock, Edward Cutrell, Susumu Harada, and I. Scott MacKenzie

  45. [54]

    Zeshi Yang and Zhiqi Yin. 2021. Efficient hyperparameter optimization for physics-based character animation. Proceedings of the ACM on Computer Graphics and Interactive Techniques 4, 1, Article 11 (April 2021). https://doi.org/10.1145/ 3451254

    2021

  46. [55]

    Wobbrock

    Jacob O. Wobbrock. 2007. Measures of text entry performance. In Text en- try systems: Mobility, accessibility, universality, I. Scott MacKenzie and Kumiko Tanaka-Ishii (Eds.). San Francisco: Morgan Kaufmann, 47–74

    2007

  47. [56]

    Shumin Zhai, Michael Hunter, and Barton A. Smith. 2000. The Metropolis keyboard – an exploration of quantitative techniques for virtual keyboard design. In UIST ’00: Proceedings of the 13th Annual ACM Symposium on User Interface Software and Technology. ACM, 119–128

    2000

  48. [57]

    Shumin Zhai and Per-Ola Kristensson. 2003. Shorthand writing on stylus key- board. In CHI ’03: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems. 97–104

    2003

  49. [58]

    Wobbrock

    Mingrui Ray Zhang, He Wen, and Jacob O. Wobbrock. 2019. Type, then correct: Intelligent text correction techniques for mobile text entry using neural networks. In Proceedings of the 32nd Annual ACM Symposium on User Interface Software and Technology. 843–855

    2019

  50. [59]

    Shumin Zhai. 2004. Characterizing computer input with Fitts’ law parameters – the information and non-information aspects of pointing. International Journal of Human–Computer Studies 61, 6 (2004), 791–809

    2004

  51. [2008]

    In CHI ’08: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems

    An error model for pointing based on Fitts’ law. In CHI ’08: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems . 1613–1622

  52. [2014]

    InCHI ’14: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems

    Uncertain text entry on mobile devices. InCHI ’14: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems . 2307–2316

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.