Analyses for linking language with psychological factors or behaviors predominately treat linguistic features as a static set, working with a single document per person or aggregating across multiple posts (e.g. on social media) into a single set of features. This limits language to mostly shed light on between-person differences rather than changes in behavior within-person. Here, we collected a novel dataset of daily surveys where participants were asked to describe their experienced well-being and report the number of alcoholic beverages they had within the past 24 hours. Through this data, we first build a multi-level forecasting model that is able to capture within-person change and leverage both the psychological features of the person and daily well-being responses. Then, we propose a longitudinal version of differential language analysis that finds patterns associated with drinking more (e.g. social events) and less (e.g. task-oriented), as well as distinguishing patterns of heavy drinks versus light drinkers.

Social science NLP tasks, such as emotion or humor detection, are required to capture the semantics along with the implicit pragmatics from text, often with limited amounts of training data. Instruction tuning has been shown to improve the many capabilities of large language models (LLMs) such as commonsense reasoning, reading comprehension, and computer programming. However, little is known about the effectiveness of instruction tuning on the social domain where implicit pragmatic cues are often needed to be captured. We explore the use of instruction tuning for social science NLP tasks and introduce Socialite-Llama — an open-source, instruction-tuned Llama. On a suite of 20 social science tasks, Socialite-Llama improves upon the performance of Llama as well as matches or improves upon the performance of a state-of-the-art, multi-task finetuned model on a majority of them. Further, Socialite-Llama also leads to improvement on 5 out of 6 related social tasks as compared to Llama, suggesting instruction tuning can lead to generalized social understanding. All resources including our code, model and dataset can be found through [bit.ly/socialitellama](https://bit.ly/socialitellama/).

Anxiety disorders are the most common of mental illnesses, but relatively little is known about how to detect them from language. The primary clinical manifestation of anxiety is worry associated cognitive distortions, which are likely expressed at the discourse-level of semantics. Here, we investigate the development of a modern linguistic assessment for degree of anxiety, specifically evaluating the utility of discourse-level information in addition to lexical-level large language model embeddings. We find that a combined lexico-discourse model outperforms models based solely on state-of-the-art contextual embeddings (RoBERTa), with discourse-level representations derived from Sentence-BERT and DiscRE both providing additional predictive power not captured by lexical-level representations. Interpreting the model, we find that discourse patterns of causal explanations, among others, were used significantly more by those scoring high in anxiety, dovetailing with psychological literature.

Psychological states unfold dynamically; to understand and measure mental health at scale we need to detect and measure these changes from sequences of online posts. We evaluate two approaches to capturing psychological changes in text: the first relies on computing the difference between the embedding of a message with the one that precedes it, the second relies on a “human-aware” multi-level recurrent transformer (HaRT). The mood changes of timeline posts of users were annotated into three classes, ‘ordinary,’ ‘switching’ (positive to negative or vice versa) and ‘escalations’ (increasing in intensity). For classifying these mood changes, the difference-between-embeddings technique – applied to RoBERTa embeddings – showed the highest overall F1 score (0.61) across the three different classes on the test set. The technique particularly outperformed the HaRT transformer (and other baselines) in the detection of switches (F1 = .33) and escalations (F1 = .61).Consistent with the literature, the language use patterns associated with mental-health related constructs in prior work (including depression, stress, anger and anxiety) predicted both mood switches and escalations.

Natural language is generated by people, yet traditional language modeling views words or documents as if generated independently. Here, we propose human language modeling (HuLM), a hierarchical extension to the language modeling problem where by a human- level exists to connect sequences of documents (e.g. social media messages) and capture the notion that human language is moderated by changing human states. We introduce, HaRT, a large-scale transformer model for solving HuLM, pre-trained on approximately 100,000 social media users, and demonstrate it’s effectiveness in terms of both language modeling (perplexity) for social media and fine-tuning for 4 downstream tasks spanning document- and user-levels. Results on all tasks meet or surpass the current state-of-the-art.

Many recent works in natural language processing have demonstrated ability to assess aspects of mental health from personal discourse. At the same time, pre-trained contextual word embedding models have grown to dominate much of NLP but little is known empirically on how to best apply them for mental health assessment. Using degree of depression as a case study, we do an empirical analysis on which off-the-shelf language model, individual layers, and combinations of layers seem most promising when applied to human-level NLP tasks. Notably, we find RoBERTa most effective and, despite the standard in past work suggesting the second-to-last or concatenation of the last 4 layers, we find layer 19 (sixth-to last) is at least as good as layer 23 when using 1 layer. Further, when using multiple layers, distributing them across the second half (i.e. Layers 12+), rather than last 4, of the 24 layers yielded the most accurate results.

In human-level NLP tasks, such as predicting mental health, personality, or demographics, the number of observations is often smaller than the standard 768+ hidden state sizes of each layer within modern transformer-based language models, limiting the ability to effectively leverage transformers. Here, we provide a systematic study on the role of dimension reduction methods (principal components analysis, factorization techniques, or multi-layer auto-encoders) as well as the dimensionality of embedding vectors and sample sizes as a function of predictive performance. We first find that fine-tuning large models with a limited amount of data pose a significant difficulty which can be overcome with a pre-trained dimension reduction regime. RoBERTa consistently achieves top performance in human-level tasks, with PCA giving benefit over other reduction methods in better handling users that write longer texts. Finally, we observe that a majority of the tasks achieve results comparable to the best performance with just 1/12 of the embedding dimensions.

Much of natural language processing is focused on leveraging large capacity language models, typically trained over single messages with a task of predicting one or more tokens. However, modeling human language at higher-levels of context (i.e., sequences of messages) is under-explored. In stance detection and other social media tasks where the goal is to predict an attribute of a message, we have contextual data that is loosely semantically connected by authorship. Here, we introduce Message-Level Transformer (MeLT) – a hierarchical message-encoder pre-trained over Twitter and applied to the task of stance prediction. We focus on stance prediction as a task benefiting from knowing the context of the message (i.e., the sequence of previous messages). The model is trained using a variant of masked-language modeling; where instead of predicting tokens, it seeks to generate an entire masked (aggregated) message vector via reconstruction loss. We find that applying this pre-trained masked message-level transformer to the downstream task of stance detection achieves F1 performance of 67%.

Human natural language is mentioned at a specific point in time while human emotions change over time. While much work has established a strong link between language use and emotional states, few have attempted to model emotional language in time. Here, we introduce the task of affective language forecasting – predicting future change in language based on past changes of language, a task with real-world applications such as treating mental health or forecasting trends in consumer confidence. We establish some of the fundamental autoregressive characteristics of the task (necessary history size, static versus dynamic length, varying time-step resolutions) and then build on popular sequence models for words to instead model sequences of language-based emotion in time. Over a novel Twitter dataset of 1,900 users and weekly + daily scores for 6 emotions and 2 additional linguistic attributes, we find a novel dual-sequence GRU model with decayed hidden states achieves best results (r = .66) significantly out-predicting, e.g., a moving averaging based on the past time-steps (r = .49). We make our anonymized dataset as well as task setup and evaluation code available for others to build on.

Mental health predictive systems typically model language as if from a single context (e.g. Twitter posts, status updates, or forum posts) and often limited to a single level of analysis (e.g. either the message-level or user-level). Here, we bring these pieces together to explore the use of open-vocabulary (BERT embeddings, topics) and theoretical features (emotional expression lexica, personality) for the task of suicide risk assessment on support forums (the CLPsych-2019 Shared Task). We used dual context based approaches (modeling content from suicide forums separate from other content), built over both traditional ML models as well as a novel dual RNN architecture with user-factor adaptation. We find that while affect from the suicide context distinguishes with no-risk from those with “any-risk”, personality factors from the non-suicide contexts provide distinction of the levels of risk: low, medium, and high risk. Within the shared task, our dual-context approach (listed as SBU-HLAB in the official results) achieved state-of-the-art performance predicting suicide risk using a combination of suicide-context and non-suicide posts (Task B), achieving an F1 score of 0.50 over hidden test set labels.