Abstract: The System Usability Scale (SUS), developed by Brooke (Usability evaluation in industry, Taylor & Francis, London, pp 189-194, 1996), had a great success among usability practitioners since it is a quick and easy to use measure for collecting users' usability evaluation of a system. Recently, Lewis and Sauro (Proceedings of the human computer interaction international conference (HCII 2009), San Diego CA, USA, 2009) have proposed a two-factor structure-Usability (8 items) and Learnability (2 items)-suggesting that practitioners might take advantage of these new factors to extract additional information from SUS data. In order to verify the dimensionality in the SUS' two-component structure, we estimated the parameters and tested with a structural equation model the SUS structure on a sample of 196 university users. Our data indicated that both the unidimensional model and the two-factor model with uncorrelated factors proposed by Lewis and Sauro (Proceedings of the human computer interaction international conference (HCII 2009), San Diego CA, USA, 2009) had a not satisfactory fit to the data. We thus released the hypothesis that Usability and Learnability are independent components of SUS ratings and tested a less restrictive model with correlated factors. This model not only yielded a good fit to the data, but it was also significantly more appropriate to represent the structure of SUS ratings.

Abstract: This paper argues that exceptions and other instances of morpheme-specific phonology are best analyzed in Optimality Theory (OT) in terms of lexically indexed markedness and faithfulness constraints. This approach is shown to capture locality restrictions, distinctions between exceptional and truly impossible patterns, distinctions between blocking and triggering, and distinctions between variation and exceptionality. It is contrasted with other OT analyses of exceptions, in particular those that disallow lexically indexed markedness constraints and those that invoke lexically specified rankings (that is, cophonologies). The data discussed are from Assamese, Finnish and Yine (formerly Piro). A learnability account of the genesis of lexically indexed constraints is also provided, in which indexation is used to resolve inconsistency detected by Tesar and Smolensky's (1998, 2000) Recursive Constraint Demotion algorithm.

Abstract: A key question in the debate on conversation analysis as an approach to SLA concerns the role of cognition in interaction and learning. Where is cognition located, and how is understanding in interaction achieved? For an empirically grounded answer, I will explore the procedural apparatus that sustains socially shared cognition. Following a brief introduction of three discursive approaches to cognition as socially shared, the article will examine how interactional organizations and linguistic resources serve to generate and sustain mutual understanding in a segment of ordinary conversation between an LI speaker and an L2 speaker of English. I will then discuss the standard treatment of repair in interactionist SLA from a conversation-analytic perspective. Lastly, I will consider how interactional competencies may be learnable, and how their learnability informs the issue of whether CA is capable of furnishing an explication of second language learning without the help of exogenous theory.

Abstract: Morphological concatenation often triggers phonological processes For instance, addition of the plural suffix /-ən/ to Dutch nouns causes vowel lengthening in some nouns due to the stress-to-weight principle ([xɑt] vs [ˈxaːtən] ‘hole’) These kinds of processes often apply only to a subset of words – not all Dutch nouns undergo this process ([kɑt] vs [ˈkɑtən] ‘cat’) Nouns need to be lexically indexed as either undergoing this process or not I investigate how phonological grammar and lexical indexation are learned when learners are confronted with data like these Based on learnability considerations, I hypothesise that learners acquire a grammar with default non-alternation, so that novel items are treated as non-alternating I report the results of artificial language learning experiments compatible with this hypothesis, and model these results in a version of the Biased Constraint Demotion algorithm (Prince & Tesar 2004)

Abstract: PHONOLOGICAL TRENDS IN THE LEXICON: THE ROLE OF CONSTRAINTS

Abstract: The Statistical Query model was introduced in [6] to handle noise in the well-known PAC model. In this model the learner gains information about the target concept by asking for various statistics about it. Characterizing the number of queries required by learning a given concept class under fixed distribution was already considered in [3] for weak learning; then in [8] strong learnability was also characterized. However, the proofs for these results in [3,10,8] (and for strong learnability even the characterization itself) are rather complex; our main goal is to present a simple approach that works for both problems. Additionally, we strengthen the result on strong learnability by showing that a class is learnable with polynomially many queries iff all consistent algorithms use polynomially many queries, and by showing that proper and improper learning are basically equivalent. As an example, we apply our results on conjunctions under the uniform distribution.

Abstract: Iterated learning in populations of Bayesian agents Kenny Smith (kenny.smith@northumbria.ac.uk) Cognition and Communication Research Centre, Division of Psychology, Northumbria University Northumberland Building,Northumberland Road,Newcastle-upon-Tyne, NE1 8ST,UK Abstract languages during their transmission. The classic example of this second constraint is the mismatch between the infinite ex- pressivity of languages and the finite set of data from which such languages must be learned. This transmission bottleneck favours languages which can be recreated from a subset via generalisation. Recursive compositionality is one such gen- eralisation (e.g. Kirby, 2002; Brighton, 2002), and therefore represents an adaptation by language in response to pressure arising from transmission factors external to the human mind. While this evolutionary process requires certain learner biases (e.g. ability to generalise), it does not arise as a consequence of these learning biases alone, but is modulated by the trans- mission bottleneck (Brighton, Smith, & Kirby, 2005). This suggests that the biases of language learners can’t simply be read off from typological distributions. However, this transmission-mediated view of the relation- ship between learning biases and typology has recently been thrown into doubt by some modelling work in the Bayesian framework. As discussed below, Griffiths and Kalish (2007) show that iterated learning in populations of Bayesian learn- ers produces outcomes which are solely determined by the biases of language learners: in other words, in the linguis- tic case, the relationship between learning bias and language typology might be a transparent one after all. It is shown here that a variant of Griffiths & Kalish’s model (where each learner selects a single grammar after observing data produced by multiple individuals, rather than a single in- dividual) leads to a blurring of the relationship between prior biases of learners and outcomes of cultural evolution: popu- lations of Bayesian agents converge on distributions of lan- guages which are dependent on both the biases of language learners and transmission factors (such as the diversity of models a learner is exposed to). Previous analytic results (Griffiths & Kalish, 2007) show that repeated learning and transmission of languages in populations of Bayesian learners results in distributions of languages which directly reflect the biases of learners. This result potentially has profound implications for our understanding of the link between the human language learning apparatus and the dis- tribution of languages in the world. It is shown here that a variation on these models (such that learners learn from the linguistic behaviour of multiple individuals, rather than a sin- gle individual) changes this transparent relationship between learning bias and typology. This suggests that inferring learn- ing bias from typology (or population behaviour from labora- tory diffusion chains) is potentially unsafe. Keywords: language learning; iterated learning; Bayesian learning; cultural evolution; language universals Introduction What is the relationship between the biases of language learn- ers and the observed distribution of languages in the world? Under the standard generative account (e.g. Chomsky, 1965), a direct mapping is assumed between the mental apparatus of language learners and language structure. In the strongest possible form (e.g. Baker, 2001), the claim is that we can read off the structure of the language faculty from the typological distribution of languages in the world. A second account which posits a similarly close match be- tween the biases of language learners and the structure of lan- guage arises from considerations of cultural evolution (Chris- tiansen & Chater, 2008). Rather than language structure being strongly constrained by a highly restrictive domain- specific learning apparatus, the idea is that languages have adapted over repeated episodes of learning and production in response to much weaker (and possibly domain-general) constraints arising from the biases of language learners. This process is sometimes called iterated learning: the outcome of learning at one generation provides the input to learning at the next. While typologically unattested languages might be both possible and even learnable, the languages we see in the world will typically be selected from the restricted set of highly learnable languages: languages which are hard to learn will tend to change, and those which are easy to learn will be preserved, eventually yielding languages which are uniformly well-fitted to the biases of language learners. We have previ- ously termed this evolutionary pressure cultural selection for learnability (Brighton, Kirby, & Smith, 2005). Are learner biases the only factor shaping the distribu- tion of languages in the world? It has been argued (see e.g. Kirby, 2002; Zuidema, 2003; Brighton, Kirby, & Smith, 2005; Kirby, Dowman, & Griffiths, 2007) that, at a minimum, language must be seen as a compromise between two fac- tors: the biases of learners, and other constraints acting on Summary of iterated learning results for Bayesian learners Bayesian learners select a hypothesis h according to its pos- terior probability in light of some data d: P(h|d) = P(d|h)P(h) ∑ h P(d|h)P(h) P(d|h) gives the likelihood of data d being produced un- der hypothesis h, and P(h) gives the prior probability of each hypothesis. For models of iterated learning of language, the set of hypotheses are interpreted as the set of possible gram- mars, data are sets of utterances from which learners must induce a language, and the prior probability distribution over grammars arises from the bias (domain-specific or domain- general, innate or learned) of learners.

Abstract: This paper proposes solutions to two semantic learnability problems that have featured prominently in the literature on language acquisition. Both problems have often been deemed unsolvable for language learners as a matter of logic, and they have accordingly been taken to motivate principles making sure they will not actually arise in the course of language acquisition. One problem concerns the acquisition of ambiguous sentences whose readings are related by entailment. Crain et al.’s (1994) Semantic Subset Principle is intended to preempt the problem by preventing acquisition of the weaker reading before the stronger reading has been acquired. In contrast, we demonstrate that this very order of acquisition becomes feasible in principle if children can exploit non-truth-conditional evidence of various kinds or evidence from sentences containing downward entailing operators. The other learnability problem concerns the potential need for expunction of certain readings of ambiguous sentences from a child’s grammar. It has often been assumed that, in the absence of negative evidence, such expunction is impossible, and Wexler and Manzini (1987) posit a Subset Principle to preempt the problematic learning scenario. We argue, however, that if the evidence available to the child includes dialogues, and if listeners are expected to interpret speakers’ utterances charitably, then expunction of unavailable readings is possible in principle.

Abstract: This paper describes evaluating interactive entertainment by understanding embodied learning in games, which is a perspective that situates the learning that a player must go through to play a game in a skill-based environment. Our goal was to arrive at a tool for designers to improve learnability from this perspective. To study embodied learning, we use the concept of breakdown, which happens when our experience fails to aid our everyday actions and decision-making. We conducted a study to investigate learning in games from which we constructed a framework of 17 patterns of breakdown and a set of guidelines to aid heuristic evaluation of video games and to help designers support breakdown in interactions, which support players’ learning, so that they do not become breakdowns in illusion, which break players’ immersion [19]. Author Keywords video games, user experience, learnability, embodied interaction, flow, immersion, entertainment

Abstract: We establish that stability is necessary and sufficient for learning, even in the General Learning Setting where uniform convergence conditions are not necessary for learning, and where learning might only be possible with a non-ERM learning rule. This goes beyond previous work on the relationship between stability and learnability, which focused on supervised classification and regression, where learnability is equivalent to uniform convergence and it is enough to consider the ERM.

Abstract: The majority of the extant languages have one of three dominant basic word orders: SVO, SOV or VSO. Various hypotheses have been proposed to explain this word order bias, including the existence of a universal grammar, the learnability imposed by cognitive constraints, the descent of modern languages from an ancestral protolanguage, and the constraints from functional principles. We run simulations using a multi-agent computational model to study this bias. Following a local order approach, the model simulates individual language processing mechanisms in production and comprehension. The simulation results demonstrate that the semantic structures that a language encodes can constrain the global syntax, and that local syntax can help trigger bias towards the global order SOV/SVO (or VOS/OVS).

Abstract: In the recent literature on monetary and fiscal policy design, adoption of policies that induce both determinacy and learnability of equilibrium has been considered fundamental to economic stabilization. We study the connections between determinacy of rational expectations equilibrium, and expectational stability or learnability of that equilibrium, in a general class of purely forward-looking models. We ask what types of economic assumptions drive differences in the necessary and sufficient conditions for the two criteria. We apply our result to a relatively general New Keynesian model. Our framework is sufficiently flexible to encompass lags in information, a cost channel for monetary policy, and either Euler equation or infinite horizon approaches to learning. We are able to isolate conditions under which determinacy does and does not imply learnability, and also conditions under which long horizon forecasts make a clear difference to conclusions about expectational stability. The sharpest result is that informational delays break equivalence connections between determinacy and learnability.

Abstract: Can second language (L2) learners acquire a grammar that allows a subset of the structures allowed by their native grammar? This question is addressed here with respect to acquisition of phonotactics. On the assumption that the L2 initial state equals the native grammar's final state, learnability theory would predict that a lack of negative evidence for phonotactic structures that are illegal in the target language precludes acquisition of the target grammar. This prediction is tested for L1-Russian (superset) and L1-Spanish (subset) L2 learners of Dutch by means of word-likeness judgments and lexical decision experiments. Participants responded to nonwords containing consonant clusters in onsets and codas that are legal (1) only in Russian, (2) only in Russian and Dutch, or (3) in all three languages. The results converge to show that advanced L1-Russian and L1-Spanish L2 learners possess native-like phonotactic knowledge. Analysis shows that this knowledge cannot be attributed to transfer of lexical st...

Abstract: A classic debate in cognitive science revolves around understanding how children learn complex linguistic rules, such as those governing restrictions on verb alternations, without negative evidence. Traditionally, formal learnability arguments have been used to claim that such learning is impossible without the aid of innate language-specific knowledge. However, recently, researchers have shown that statistical models are capable of learning complex rules from only positive evidence. These two kinds of learnability analyses differ in their assumptions about the distribution from which linguistic input is generated. The former analyses assume that learners seek to identify grammatical sentences in a way that is robust to the distribution from which the sentences are generated, analogous to discriminative approaches in machine learning. The latter assume that learners are trying to estimate a generative model, with sentences being sampled from that model. We show that these two learning approaches differ in their use of implicit negative evidence – the absence of a sentence – when learning verb alternations, and demonstrate that human learners can produce results consistent with the predictions of both approaches, depending on how the learning problem is presented.

Abstract: Mobile usability is not one-dimensional property of user interface, it has many components and attributes [1] The most attributes associated with the mobile usability are as following: Satisfaction, Efficiency, Learnability, Lack of Errors, and Memorability The assessment and importance of the usability attributes varies based on the application and the type of the users. For the expert usersâ?? lack of errors, reliability and efficiency are important and for the novice users beside mentioned factors the adjustability, learnability and memorability are essentials factors. This paper reveals different usability considerations for mobile learning system with approparte guidelines.

Abstract: Mobile devices have the potential to support many older adults (age 65+) in their daily lives. However, older adults find it difficult to learn to use many existing mobile device applications and their interfaces. The goal of this dissertation research is to improve the learnability of mobile software user interfaces for older adults. To achieve this goal, we will investigate three complementary design approaches that have not been well explored for this population.

Abstract: In optimality theory (OT) the essence of both language learning in general (learnability) and language acquisition (the actual development children go through) entails the ranking of constraints from an initial state of the grammar to the language-specific ranking of the target grammar. This is the common denominator in all OT studies on language acquisition and learning. There are many unsettled issues, however. Are the constraints innate or do they emerge during acquisition (nature-nurture)? And if they emerge, where do they come from? What is the initial state? Does the (re)ranking of constraints only involve the demotion of markedness constraints, the promotion of faithfulness constraints, or can it be achieved by both the demotion and the promotion of constraints? Another issue is whether comprehension and production are mediated by the same grammar or whether there is one grammar for comprehension and another for pro!

Abstract: Given a constraint set with k constraints in the framework of Optimality Theory (OT), what is its capacity as a classification scheme for linguistic data? One useful measure of this capacity is the size of the largest data set of which each subset is consistent with a different grammar hypothesis. This measure is known as the Vapnik-Chervonenkis dimension (VCD) and is a standard complexity measure for concept classes in computational learnability theory. In this work, I use the three-valued logic of Elementary Ranking Conditions to show that the VCD of Optimality Theory with k constraints is k-1. Analysis of OT in terms of the VCD establishes that the complexity of OT is a well-behaved function of k and that the 'hardness' of learning in OT is linear in k for a variety of frameworks that employ probabilistic definitions of learnability.

Abstract: In this paper we are concerned with some general properties of scientific hypotheses. We investigate the relationship between the situation when the task is to verify a given hypothesis, and when a scientist has to pick a correct hypothesis from an arbitrary class of alternatives. Both these procedures are based on induction. We understand hypotheses as generalized quantifiers of types $\left\langle 1\right\rangle$ or $\left\langle 1,1\right\rangle$. Some of their formal features, like monotonicity, appear to be of great relevance. We first focus on monotonicity, extendability and persistence of quantifiers. They are investigated in context of epistemological verifiability of scientific hypotheses. In the second part we show that some of these properties imply learnability. As a result two strong paradigms are joined: the paradigm of computational epistemology (see e.g.[6,5] ), which goes back to the notion of identification in the limit as formulated in [4], and the paradigm of investigating natural language determiners in terms of generalized quantifiers in finite models (see e.g.[1]).

Abstract: In the past twenty years the connectionist approach to language development and learning has emerged as an alternative,e to traditional linguistic theories. This article introduces the connectionist paradigm by describing basic operating principles of neural network models as it;ell as different network architectures. The application of neural network models to explanations for linguistic problems is illustrated by reviewing a number of models for different aspects of language development, from speech sound acquisition to the development of syntax. Two main ben(fits of the connectionist approach are highlighted: implemented models offer a high degree of specificity, for a particular theory, and the explicit integration of a learning process into theory building allows for detailed investigation of the effect of he linguistic environment on a child. Issues regarding learnability or the need to assume innate and domain specific knowledge thus become an empirical question that can be answered by evaluating a model's performance.

Abstract: This commentary addresses the relevance of detectability to a theory of learning uninterpretable features in the second language (L2). Detectability of features is illustrated in an application of Signal Detection Theory. By analogy with development of phonemic categories in the first language (L1), the notion of paring down the repertoire of uninterpretable features is considered.

Abstract: The wh-marking of questions in child English is as early as the appearance of the wh-questions themselves. The wh-marking of questions in child Dutch (and the other Germanic languages) is delayed until the acquisition of articles and free anaphoric pronouns. An acquisition procedure is proposed that succeeds to set first a typological difference, V2 for Dutch and SVfinO for English. The different setting of the typological parameters determines the wh-development in subsequent acquisition steps. The learnability approach relativizes Chomsky’s poverty of the stimulus, but affirms his position that language is ‘perfect’ in the sense of being learnable as a cultural construct without the assumption of innate grammar-specific a prioris.

Abstract: In order to systematize existing results, we propose to analyze the learnability of boolean functions computed by an algebraically defined model, programs over monoids. The expressiveness of the model, hence its learning complexity, depends on the algebraic structure of the chosen monoid. We identify three classes of monoids that can be identified, respectively, from Membership queries alone, Equivalence queries alone, and both types of queries. The algorithms for the first class are new to our knowledge, while those for the other two are combinations or particular cases of known algorithms. Learnability of these three classes captures many previous learning results. Moreover, by using nontrivial taxonomies of monoids, we can argue that using the same techniques to learn larger classes of boolean functions seems to require proving new circuit lower bounds or proving learnability of DNF formulas.

Abstract: A classic debate in cognitive science revolves around understanding how children learn complex linguistic rules, such as those governing restrictions on verb alternations, without negative evidence. Traditionally, formal learnability arguments have been used to claim that such learning is impossible without the aid of innate language-specific knowledge. However, recently, researchers have shown that statistical models are capable of learning complex rules from only positive evidence. These two kinds of learnability analyses differ in their assumptions about the distribution from which linguistic input is generated. The former analyses assume that learners seek to identify grammatical sentences in a way that is robust to the distribution from which the sentences are generated, analogous to discriminative approaches in machine learning. The latter assume that learners are trying to estimate a generative model, with sentences being sampled from that model. We show that these two learning approaches differ in their use of implicit negative evidence - the absence of a sentence - when learning verb alternations, and demonstrate that human learners can produce results consistent with the predictions of both approaches, depending on how the learning problem is presented.

Abstract: The design and evaluation of haptic icons – brief, meaningful tactile stimuli – has been studied extensively in the research community. Haptic icons are designed to support communication of information through the often-underutilized haptic modality. However, the learnability of haptic icons has not been evaluated in an ecologically plausible, longitudinal deployment scenario. This thesis endeavours to evaluate the learnability of haptic icons in a realistic context. We assign abstract meanings based on a realistic context to a large, previously developed set of rhythmic haptic stimuli. Then, during a period of 12 sessions over 4 weeks, we train users to recognize these icons and observe identification performance under workload using a Tetris game interruption task. Icons are presented to users in sets of 7. Upon the mastery of their current 7 icons, the user graduates to a new set, but must remember previously learned icons. We discover that perceptual discriminability dominates learnability – the semantics of the icons have very little effect. We also find evidence that design based on multidimensional scaling (MDS) is adequate for developing haptic stimulus sets, but can be quite conservative in its identification performance predictions during deployment. Haptic icon learning is characterized by a peak in difficulty after learning progresses past a single group 7 icons, which may be explained by cognitive long-term encoding and an increase in perceptual sensitivity. In addition, we present a series of heuristics for designing rhythmic haptic icons, as well as guidelines for haptic icon training and advice for hardware designers. In an attempt to increase the expressiveness and learnability of rhythmic haptic icons, we explore the addition of melody. We iteratively develop a second set of 30 melodic haptic icons using an MDS methodology. We discover that rhythm dominates user categorization of melodies. This work also results in a set of heuristics for designing melodic icons. Finally, we evaluate the learnability of this new melodic set using our previous longitudinal methodology. Our results indicate that purely rhythmic haptic icons are easier to learn than melodic haptic icons that are grouped by rhythm and are thus more viable for deployment.