Episodic-like memory in wild free-living blue tits and great tits.

Episodic-like memory in non-human animals represents the behavioral characteristics of human episodic memory-the ability to mentally travel backward in time to "re-live" past experiences. A focus on traditional model species of episodic-like memory may overlook taxa possessing this cognitive ability and consequently its evolution across species. Experiments conducted in the wild have the potential to broaden the scope of episodic-like memory research under the natural conditions in which they evolved. We combine two distinct yet complementary episodic-like memory tasks (the what-where-when memory and incidental encoding paradigms), each targeting a different aspect of human episodic memory, namely the content (what-where-when) and process (incidental encoding), to comprehensively test the memory abilities of wild, free-living, non-caching blue tits (Cyanistes caeruleus) and great tits (Parus major). Automated feeders with custom-built programs allowed for experimental manipulation of spatiotemporal experiences on an individual-level basis. In the what-where-when memory experiment, after learning individualized temporal feeder rules, the birds demonstrated their ability to recall the "what" (food type), "where" (feeder location), and "when" (time since their initial visit of the day) of previous foraging experiences. In the incidental encoding experiment, the birds showed that they were able to encode and recall incidental spatial information regarding previous foraging experiences ("where" test), and juveniles, but not adults, were also able to recall incidentally encoded visual information ("which" test). Consequently, this study presents multiple lines of converging evidence for episodic-like memory in a wild population of generalist foragers, suggesting that episodic-like memory may be more taxonomically widespread than previously assumed.

Improving young patient - practitioner communication about sexual health during medical visits.

OBJECTIVES: Young people face barriers that lead to gaps in sexual and reproductive health care communications. Issues such as discomfort discussing sexual health lead to inadequate delivery of services resulting in unintended pregnancies and STIs. Closing this communication gap between patients and health care practitioners would improve communication and health outcomes. The objective of this study was to gain feedback from focus groups about: (a) barriers and facilitators to communication surrounding sexual health and (b) the feasibility and acceptability of a question prompt list (QPL) and informational video emphasizing asking questions about sexual health during medical visits as tools young people could use to be more involved during visits. METHODS: Three focus groups were conducted: two with young adults (n=14) and one with practitioners (n=5) of sexual/reproductive health care services. Practitioners were recruited from healthcare clinics. RESULTS: Young adults were aged 18-22 years old. Participants identified barriers to communication such as embarrassment over sexual health topics and practitioner assumptions about patients' base of knowledge. A facilitator to communication was patient-friendly language. Focus group participants offered suggestions on how to improve the QPL as well as themes that should be covered in an educational video. Participants viewed the QPL and educational video as useful for encouraging conversations between patient and practitioner. CONCLUSIONS: Many barriers obstruct communication between young adults and practitioners on topics relating to sexual health. Both the QPL and an educational video could be used to enhance patient-practitioner communication.

FLMatchQA: a recursive neural network-based question answering with customized federated learning model.

More sophisticated data access is possible with artificial intelligence (AI) techniques such as question answering (QA), but regulations and privacy concerns have limited their use. Federated learning (FL) deals with these problems, and QA is a viable substitute for AI. The utilization of hierarchical FL systems is examined in this research, along with an ideal method for developing client-specific adapters. The User Modified Hierarchical Federated Learning Model (UMHFLM) selects local models for users' tasks. The article suggests employing recurrent neural network (RNN) as a neural network (NN) technique for learning automatically and categorizing questions based on natural language into the appropriate templates. Together, local and global models are developed, with the worldwide model influencing local models, which are, in turn, combined for personalization. The method is applied in natural language processing pipelines for phrase matching employing template exact match, segmentation, and answer type detection. The (SQuAD-2.0), a DL-based QA method for acquiring knowledge of complicated SPARQL test questions and their accompanying SPARQL queries across the DBpedia dataset, was used to train and assess the model. The SQuAD2.0 datasets evaluate the model, which identifies 38 distinct templates. Considering the top two most likely templates, the RNN model achieves template classification accuracy of 92.8% and 61.8% on the SQuAD2.0 and QALD-7 datasets. A study on data scarcity among participants found that FL Match outperformed BERT significantly. A MAP margin of 2.60% exists between BERT and FL Match at a 100% data ratio and an MRR margin of 7.23% at a 20% data ratio.

Understanding consumer acceptability of verjuice, its potential applications and sensory and chemical drivers of liking.

The wine sector is working to add value, enhance sustainability and reduce waste, yet often creating new products with unknown consumer acceptance. Verjuice, juice made from discarded unripe thinned grapes, is an example. Whilst verjuice has various culinary uses, its versatility in beverages continues to expand. However, its sensory drivers of liking when consumed as a drink, and their potential impact on its application remain unclear. Chemical drivers of sensory characteristics are also unknown representing a critical knowledge gap needed to guide product innovation. This study aimed to provide new knowledge regarding consumer acceptability of verjuice by identifying its sensory drivers of liking as a beverage, evaluating its potential use in different applications and identifying chemical drivers of its sensory characteristics. New Zealand consumers (n = 93) evaluated 13 verjuice samples from different countries. Furthermore, verjuice familiarity and its influence on emotional response was investigated to determine likely future consumer engagement. Sweetness was the most positive driver of liking in verjuice overall, followed by fruity and floral notes, smoothness, and to a lesser extent citrus flavour. Consumers expressed varied preferences for verjuice's sensory profile, with fruity, floral, and honey flavours driving beverage application, whilst winey and green apple notes were more associated with culinary scenarios. Some association between chemical parameters and sensory attributes were evident (e.g. sugars were highly associated with perceived sweetness, fruity, and floral attributes; these attributes also shared proximity with 1-hexanol and cyclohexanol). The general idea of verjuice elicited positive valence emotions, but consumers who were 'not familiar' felt more curious, and those 'familiar' felt happier and more satisfied. Findings highlight the potential to tailor verjuice for specific applications by understanding desired sensory profiles and related chemical parameters. Recognising the interplay between familiarity level and emotional response is crucial for positioning the product in the marketplace and fostering consumer engagement. Marketing initiatives are needed to increase verjuice familiarity and support product innovation, leading to increased product appeal.

[A medical visual question answering approach based on co-attention networks].

Recent studies have introduced attention models for medical visual question answering (MVQA). In medical research, not only is the modeling of "visual attention" crucial, but the modeling of "question attention" is equally significant. To facilitate bidirectional reasoning in the attention processes involving medical images and questions, a new MVQA architecture, named MCAN, has been proposed. This architecture incorporated a cross-modal co-attention network, FCAF, which identifies key words in questions and principal parts in images. Through a meta-learning channel attention module (MLCA), weights were adaptively assigned to each word and region, reflecting the model's focus on specific words and regions during reasoning. Additionally, this study specially designed and developed a medical domain-specific word embedding model, Med-GloVe, to further enhance the model's accuracy and practical value. Experimental results indicated that MCAN proposed in this study improved the accuracy by 7.7% on free-form questions in the Path-VQA dataset, and by 4.4% on closed-form questions in the VQA-RAD dataset, which effectively improves the accuracy of the medical vision question answer.

How Good Are Cardiologists at Predicting Major Adverse Events in Fontan Patients?

Background: It is unknown how well cardiologists predict which Fontan patients are at risk for major adverse events (MAEs). Objectives: The purpose of this study was to examine the accuracy of cardiologists' ability to identify the "good Fontan" patient, free from MAE within the following year, and compare that predicted risk cohort to patients who experienced MAE. Methods: This prospective, multicenter study included patients ≥10 years with lateral tunnel or extracardiac Fontan. The cardiologist was asked the yes/no "surprise" question: would you be surprised if your patient has a MAE in the next year? After 12 months, the cardiologist was surveyed to assess MAE. Agreement between cardiologist predictions of MAE and observed MAE was determined using the simple kappa coefficient. Multivariable generalized linear mixed effects models were performed to identify factors associated with MAE. Results: Overall, 146 patients were enrolled, and 99/146 (68%) patients w`ere predicted to be a "good Fontan." After 12 months, 17 (12%) experienced a MAE. The simple kappa coefficient of cardiologists' prediction was 0.17 (95% CI: 0.02-0.32), suggesting prediction of MAE was 17% better than random chance. In the multivariable cardiologist-predicted MAE (N = 47) model, diuretic/beta-blocker use (P ≤ 0.001) and systolic dysfunction (P = 0.005) were associated with MAE. In the observed multivariable MAE (N = 17) model, prior unplanned cardiac admission (P = 0.006), diuretic/beta-blocker use (P = 0.028), and ≥moderate atrioventricular valve regurgitation (P = 0.049) were associated with MAE. Conclusions: Cardiologists are marginally able to predict which Fontan patients are at risk for MAE over a year. There was overlap between factors associated with a cardiologist's prediction of risk and observed MAE, namely the use of diuretic/beta-blocker.

Large language models in psychiatry: Opportunities and challenges.

The ability of Large Language Models (LLMs) to analyze and respond to freely written text is causing increasing excitement in the field of psychiatry; the application of such models presents unique opportunities and challenges for psychiatric applications. This review article seeks to offer a comprehensive overview of LLMs in psychiatry, their model architecture, potential use cases, and clinical considerations. LLM frameworks such as ChatGPT/GPT-4 are trained on huge amounts of text data that are sometimes fine-tuned for specific tasks. This opens up a wide range of possible psychiatric applications, such as accurately predicting individual patient risk factors for specific disorders, engaging in therapeutic intervention, and analyzing therapeutic material, to name a few. However, adoption in the psychiatric setting presents many challenges, including inherent limitations and biases in LLMs, concerns about explainability and privacy, and the potential damage resulting from produced misinformation. This review covers potential opportunities and limitations and highlights potential considerations when these models are applied in a real-world psychiatric context.

Implementing Pivotal Response Treatment to Teach Question Asking to High School Students with Autism Spectrum Disorder.

The purpose of this study was to test the use of Pivotal Response Treatment (PRT) in the secondary school setting. There were two main goals: (a) to evaluate secondary education providers' ability to implement PRT with fidelity following a PRT training program; and (b) to evaluate the effects of school-implemented PRT on the social communication skills of adolescents and young adults with ASD, specifically, question-asking behavior. This concurrent multiple baseline design study across dyads investigated the use of PRT in the secondary school setting with adolescents with ASD. Specifically, it examined the impact of PRT on question-asking behavior. Education providers (n = 3) were trained to implement PRT with a secondary student with ASD. All education providers improved in their ability to use PRT strategies, though struggled with fidelity. Two students exhibited clear effects with noteworthy improvement in their use of targeted question initiations. For targeted question initiations, the weighted value for the Tau-U phase contrast between aggregated baseline and intervention phases was 0.80 and statistically significant (p < .0001). PRT is a promising approach to increasing question-asking behavior in secondary students with ASD when implemented by a trained education provider. Continued research should be a matter of priority in order to expand social skills instruction for adolescents with ASD with the hope of ultimately making a positive difference in adult outcomes.

A comparison of chain-of-thought reasoning strategies across datasets and models.

Emergent chain-of-thought (CoT) reasoning capabilities promise to improve the performance and explainability of large language models (LLMs). However, uncertainties remain about how reasoning strategies formulated for previous model generations generalize to new model generations and different datasets. In this small-scale study, we compare different reasoning strategies induced by zero-shot prompting across six recently released LLMs (davinci-002, davinci-003, GPT-3.5-turbo, GPT-4, Flan-T5-xxl and Cohere command-xlarge). We test them on six question-answering datasets that require real-world knowledge application and logical verbal reasoning, including datasets from scientific and medical domains. Our findings demonstrate that while some variations in effectiveness occur, gains from CoT reasoning strategies remain robust across different models and datasets. GPT-4 benefits the most from current state-of-the-art reasoning strategies and performs best by applying a prompt previously discovered through automated discovery.

Replay of incidentally encoded novel odors in the rat.

Although events are not always known to be important when they occur, people can remember details about such incidentally encoded information using episodic memory. Sheridan et al. (2024) argued that rats replayed episodic memories of incidentally encoded information in an unexpected assessment of memory. In one task, rats reported the third-last item in an explicitly encoded list of trial-unique odors. In a second task, rats foraged in a radial maze in the absence of odors. On a critical test, rats foraged in the maze, but scented lids covered the food. Next, memory of the third-last odor was assessed. The rats correctly answered the unexpected question. Because the odors used in the critical test were the same as those used during training, automatically encoding odors for the purpose of taking an upcoming test of memory (stimulus generalization) may have been encouraged. Here, we provided an opportunity for incidental encoding of novel odors. Previously trained rats foraged in the radial maze with entirely novel odors covering the food. Next, memory of the third-last odor was assessed. The rats correctly answered the unexpected question. High accuracy when confronted with novel odors provides evidence that the rats did not automatically encode odors for the purpose of taking an upcoming test, ruling out stimulus generalization. We conclude that rats encode multiple pieces of putatively unimportant information, and later replayed a stream of novel episodic memories when that information was needed to solve an unexpected problem.

American Board of Surgery In-Training Exam Performance Predicted by Question Bank Use While Unassociated With Other Learning Strategies.

INTRODUCTION: There is no consensus regarding optimal curricula to teach cognitive elements of general surgery. The American Board of Surgery In-Training Exam (ABSITE) aims to measure trainees' progress in attaining this knowledge. Resources like question banks (QBs), Surgical Council on Resident Education (SCORE) curriculum, and didactic conferences have mixed findings related to ABSITE performance and are often evaluated in isolation. This study characterized relationships between multiple learning methods and ABSITE performance to elucidate the relative educational value of learning strategies. METHODS: Use and score of QB, SCORE use, didactic conference attendance, and ABSITE percentile score were collected at an academic general surgery residency program from 2017 to 2022. QB data were available in the years 2017-2018 and 2021-2022 during institutional subscription to the same platform. Given differences in risk of qualifying exam failure, groups of ≤30th and >30th percentile were analyzed. Linear quantile mixed regressions and generalized linear mixed models determined factors associated with ABSITE performance. RESULTS: Linear quantile mixed regressions revealed a relationship between ABSITE performance and QB questions completed (1.5 percentile per 100 questions, P < 0.001) and QB score (1.2 percentile per 1% score, P < 0.001), but not with SCORE use and didactic attendance. Performers >30th percentile had a significantly higher QB score. CONCLUSIONS: Use and score of QB had a significant relationship with ABSITE performance, while SCORE use and didactic attendance did not. Performers >30th percentile completed a median 1094 QB questions annually with a score of 65%. Results emphasize success of QB use as an active learning strategy, while passive learning methods warrant further evaluation.

A pre-visit video/question prompt list intervention to increase youth question-asking about attention deficit hyperactivity disorder during pediatric visits.

OBJECTIVE: To conduct a pragmatic randomized controlled trial to test the effectiveness of an ADHD question prompt list with video intervention to increase youth question-asking and provider education about ADHD during visits. METHODS: English-speaking youth ages 11-17 with ADHD and their caregivers were enrolled from two pediatric clinics. Youth were randomized to intervention or usual care groups. Intervention group adolescents watched the video and then completed an ADHD question prompt list before their visits. Multivariable regression was used to analyze the data. RESULTS: Twenty-one providers and 102 of their patients participated. Intervention group youth were significantly more likely to ask one or more questions about ADHD and its treatment than usual care youth (odds ratio=5.4, 95 % Confidence Interval (CI)= 1.8, 15.9). Providers were significantly more likely to educate youth who asked one or more questions during visits about more ADHD medication areas (unstandardized beta=0.98, 95 % CI=0.31 to 1.64) and more non-medication strategies for ADHD (unstandardized beta=0.50, 95 % CI=0.13 to 0.88). CONCLUSION: The intervention increased youth question-asking about ADHD and its treatment. Providers provided more education to youth who asked one or more questions about ADHD and its treatment. PRACTICE IMPLICATIONS: Providers and practices should consider having youth complete ADHD question prompt lists and watch the video before visits to increase youth question-asking during visits.

When is it appropriate to ask a question? The role of age, social context, and personality.

How do children decide when it is appropriate to ask a question? In Study 1 (preregistered), 50 4- and 5-year-olds, 50 7- and 8-year-olds, and 100 adults watched vignettes featuring a child who had a question, and participants indicated whether they thought the child should ask the question "right now." Both adults and children endorsed more question-asking to a well-known informant than to an acquaintance and to someone doing nothing than to someone busy working or busy socializing. However, younger children endorsed asking questions to someone who was busy more often than older children and adults. In addition, Big Five personality traits predicted endorsement of question-asking. In Study 2 (preregistered, N = 500), mothers' self-reports showed that children's actual question-asking varied with age, informant activity, and informant familiarity in ways that paralleled the results of Study 1. In Study 3 (N = 100), we examined mothers' responses to their children's question-asking and found that mothers' responses to their children's question-asking varied based on the mother's activity. In addition, mothers high in authoritarianism were less likely to answer their children's questions when they were busy than mothers low in authoritarianism. In sum, across three studies, we found evidence that the age-related decline in children's question-asking to their parents reflects a change in children's reasoning about when it is appropriate to ask a question.

BioInstruct: instruction tuning of large language models for biomedical natural language processing.

OBJECTIVES: To enhance the performance of large language models (LLMs) in biomedical natural language processing (BioNLP) by introducing a domain-specific instruction dataset and examining its impact when combined with multi-task learning principles. MATERIALS AND METHODS: We created the BioInstruct, comprising 25 005 instructions to instruction-tune LLMs (LLaMA 1 and 2, 7B and 13B version). The instructions were created by prompting the GPT-4 language model with 3-seed samples randomly drawn from an 80 human curated instructions. We employed Low-Rank Adaptation (LoRA) for parameter-efficient fine-tuning. We then evaluated these instruction-tuned LLMs on several BioNLP tasks, which can be grouped into 3 major categories: question answering (QA), information extraction (IE), and text generation (GEN). We also examined whether categories (eg, QA, IE, and generation) of instructions impact model performance. RESULTS AND DISCUSSION: Comparing with LLMs without instruction-tuned, our instruction-tuned LLMs demonstrated marked performance gains: 17.3% in QA on average accuracy metric, 5.7% in IE on average F1 metric, and 96% in Generation tasks on average GPT-4 score metric. Our 7B-parameter instruction-tuned LLaMA 1 model was competitive or even surpassed other LLMs in the biomedical domain that were also fine-tuned from LLaMA 1 with vast domain-specific data or a variety of tasks. Our results also show that the performance gain is significantly higher when instruction fine-tuning is conducted with closely related tasks. Our findings align with the observations of multi-task learning, suggesting the synergies between 2 tasks. CONCLUSION: The BioInstruct dataset serves as a valuable resource and instruction tuned LLMs lead to the best performing BioNLP applications.

Effect of Jellyfish Body Parts and Presentation Form on Consumers Liking, Sensory Perception, Emotions, and Food Pairings.

Although jellyfish represent a food source in Asia, limited attention has been devoted to investigating Western consumers' perception and acceptance. This study explored the role of jellyfish body parts and presentation form in determining consumer perception. A local consumer test with 106 untrained subjects (57.5% female, 18-45 years) was performed in Italy over two days on six samples of jellyfish (Rhopilema esculentum Kishinouye) differing in terms of body parts (umbrella and oral arms) and presentation form (minced, striped, and pieced). For each sample, participants expressed their overall liking and, through three check-all-that-apply tests, described their perceived sensory properties and emotions and potential preferred food pairings. The results showed a significant effect of presentation form on liking (with striped and minced samples liked more than pieced samples), 18 sensory properties, four emotions, and five food pairings. Moreover, different drivers of liking and emotions were observed for three clusters of subjects named "In favour of", "Against", and "Picky towards" eating jellyfish. In conclusion, this study found that at least one segment of consumers could accept jellyfish as novel food. Moreover, the provided results could be useful for developing innovative jellyfish-based products and dishes that meet consumers' expectations.

GastroBot: a Chinese gastrointestinal disease chatbot based on the retrieval-augmented generation.

Introduction: Large Language Models (LLMs) play a crucial role in clinical information processing, showcasing robust generalization across diverse language tasks. However, existing LLMs, despite their significance, lack optimization for clinical applications, presenting challenges in terms of illusions and interpretability. The Retrieval-Augmented Generation (RAG) model addresses these issues by providing sources for answer generation, thereby reducing errors. This study explores the application of RAG technology in clinical gastroenterology to enhance knowledge generation on gastrointestinal diseases. Methods: We fine-tuned the embedding model using a corpus consisting of 25 guidelines on gastrointestinal diseases. The fine-tuned model exhibited an 18% improvement in hit rate compared to its base model, gte-base-zh. Moreover, it outperformed OpenAI's Embedding model by 20%. Employing the RAG framework with the llama-index, we developed a Chinese gastroenterology chatbot named "GastroBot," which significantly improves answer accuracy and contextual relevance, minimizing errors and the risk of disseminating misleading information. Results: When evaluating GastroBot using the RAGAS framework, we observed a context recall rate of 95%. The faithfulness to the source, stands at 93.73%. The relevance of answers exhibits a strong correlation, reaching 92.28%. These findings highlight the effectiveness of GastroBot in providing accurate and contextually relevant information about gastrointestinal diseases. During manual assessment of GastroBot, in comparison with other models, our GastroBot model delivers a substantial amount of valuable knowledge while ensuring the completeness and consistency of the results. Discussion: Research findings suggest that incorporating the RAG method into clinical gastroenterology can enhance the accuracy and reliability of large language models. Serving as a practical implementation of this method, GastroBot has demonstrated significant enhancements in contextual comprehension and response quality. Continued exploration and refinement of the model are poised to drive forward clinical information processing and decision support in the gastroenterology field.

Assessment of Surgical Outcomes in Patients with Degenerative Cervical Myelopathy Using the 25-Question Geriatric Locomotive Function Scale: A Longitudinal Observational Study.

Introduction: Locomotive syndrome caused by degenerative musculoskeletal diseases is reported to improve with surgical treatment. However, it is unclear whether surgical treatment is effective for the locomotive syndrome developing in patients with degenerative cervical myelopathy (DCM). Thus, this study primarily aimed to longitudinally assess the change in locomotive syndrome stage before and after cervical spinal surgery for patients with DCM using the 25-question geriatric locomotive function scale (GLFS-25). A secondary objective was to identify factors associated with the postoperative improvement in the locomotive syndrome stage. Methods: We retrospectively reviewed clinical data of patients undergoing cervical spine surgery at our institution from April 2020 to May 2022 who had answered the Japanese Orthopaedic Association Cervical Myelopathy Assessment Questionnaire, visual analog scale, and GLFS-25 preoperatively and at 6 months and 1 year postoperatively. We collected demographic data, medical history, preoperative radiographic parameters, presence or absence of posterior longitudinal ligament ossification, and surgical data. Results: We enrolled 115 patients (78 men and 37 women) in the present study. Preoperatively, using the GLFS-25, 73.9% of patients had stage 3, 10.4% had stage 2, 9.6% had stage 1, 6.1% had no locomotive syndrome. The stage distribution of locomotive syndrome improved significantly at 6-months and 1-year postoperatively. The multivariable Poisson regression analysis revealed that better preoperative lower extremity function (relative risk: 3.0; 95% confidence interval: 1.01-8.8) was significantly associated with postoperative improvement in the locomotive syndrome stage. Conclusions: This is the first study to longitudinally assess the locomotive syndrome stage in patients with DCM using GLFS-25. Our results indicated that patients with DCM experienced significant improvement in the locomotive syndrome stage following cervical spine surgery. Particularly, the preoperative lower extremity function was significant in postoperative improvement in the locomotive syndrome stage.

A comprehensive analysis of the pre- and postoperative responses to each of the 12 Oxford knee score questions one year following knee arthroplasty.

AIMS: To assess the pre- and postoperative responses to each of the 12 individual Oxford Knee Score (OKS) questions and percentages of those that were better, same or worse after primary knee arthroplasty (KA). METHODS: A single centre retrospective cohort study conducted over a 24-month period which included 3259 patients with completed OKS preoperatively and 1-year after KA. There were 1286 males and 1973 females, with an overall mean age of 70.0 (range 34-94). The mean scores for each question of the OKS were compared between baseline and 1-year. The percentage of patients who reported better, the same or worse postoperative symptoms for each question were calculated and represented on a heatmap. RESULTS: There were significant (p < 0.001) improvements in all 12 questions, all of which demonstrated moderate (Q2, Q7) or large effect sizes. Improvements in individual question responses varied. Symptoms of pain and limping demonstrated the greatest improvement, with 86% of patients enjoying a positive change in their symptoms. Despite this improvement 1067 (41.4%) continued to have mild to severe pain in their knee, and 442 (17.3%) patients limped often to all the time when walking postoperatively. Whereas other questions that did not improve to the same extent for example washing and drying only improved in 53% of patients but only 347 (13.5%) had moderate/extreme trouble or found it impossible to do this postoperatively. Preoperatively four questions (Q1, Q6, Q7, Q8) demonstrated floor effects, postoperatively all questions apart from question 7 (kneeling) demonstrated ceiling effects. CONCLUSION: The mean improvement and outcome at 1-year for each of the 12 questions varied according to the patient's preoperative response. As a clinical tool, the heatmap (improvement, same and worse) will enable communication to patients about their potential change in their knee specific symptoms according to their preoperative responses. LEVEL OF EVIDENCE: Retrospective study, Level III.

Learning the Meanings of Function Words From Grounded Language Using a Visual Question Answering Model.

Interpreting a seemingly simple function word like "or," "behind," or "more" can require logical, numerical, and relational reasoning. How are such words learned by children? Prior acquisition theories have often relied on positing a foundation of innate knowledge. Yet recent neural-network-based visual question answering models apparently can learn to use function words as part of answering questions about complex visual scenes. In this paper, we study what these models learn about function words, in the hope of better understanding how the meanings of these words can be learned by both models and children. We show that recurrent models trained on visually grounded language learn gradient semantics for function words requiring spatial and numerical reasoning. Furthermore, we find that these models can learn the meanings of logical connectives and and or without any prior knowledge of logical reasoning as well as early evidence that they are sensitive to alternative expressions when interpreting language. Finally, we show that word learning difficulty is dependent on the frequency of models' input. Our findings offer proof-of-concept evidence that it is possible to learn the nuanced interpretations of function words in a visually grounded context by using non-symbolic general statistical learning algorithms, without any prior knowledge of linguistic meaning.

Question banks: credit? Or debit? A qualitative exploration of their use among medical students.

BACKGROUND: Online question banks are the most widely used education resource amongst medical students. Despite this there is an absence of literature outlining how and why they are used by students. Drawing on Deci and Ryan's self-determination theory, our study aimed to explore why and how early-stage medical students use question banks in their learning and revision strategies. METHODS: The study was conducted at Newcastle University Medical School (United Kingdom and Malaysia). Purposive, convenience and snowball sampling of year two students were employed. Ten interviews were conducted. Thematic analysis was undertaken iteratively, enabling exploration of nascent themes. Data collection ceased when no new perspectives were identified. RESULTS: Students' motivation to use question banks was predominantly driven by extrinsic motivators, with high-stakes exams and fear of failure being central. Their convenience and perceived efficiency promoted autonomy and thus motivation. Rapid feedback cycles and design features consistent with gamification were deterrents to intrinsic motivation. Potentially detrimental patterns of question bank use were evident: cueing, avoidance and memorising. Scepticism regarding veracity of question bank content was absent. CONCLUSIONS: We call on educators to provide students with guidance about potential pitfalls associated with question banks and to reflect on potential inequity of access to these resources.