In this paper, we present a novel framework which accomplishes this by exploiting latent embedding spaces to define an objective function for an LLM in an iterative RL-driven process. As an example, consider the challenge of assisting content creators in generating valuable content within a recommender ecosystem (e.g., Y ouTube, Reddit, Spotify) [Boutilier et al., 2024].

Traditional recommender systems rely on high-dimensional (latent) embeddings for modeling user-item interactions, often resulting in opaque representations that lack interpretability. Moreover, these systems offer limited control to users over their recommendations. Inspired by recent work, we introduce TExtuAl Representations for Scrutable recommendations (TEARS) to address these challenges. Instead of representing a user's interests through a latent embedding, TEARS encodes them in natural text, providing transparency and allowing users to edit them. To do so, TEARS uses a modern LLM to generate user summaries based on user preferences. We find the summaries capture user preferences uniquely. Using these summaries, we take a hybrid approach where we use an optimal transport procedure to align the summaries' representation with the learned representation of a standard VAE for collaborative filtering. We find this approach can surpass the performance of three popular VAE models while providing user-controllable recommendations. We also analyze the controllability of TEARS through three simulated user tasks to evaluate the effectiveness of a user editing its summary.

Empathy serves as a cornerstone in enabling prosocial behaviors, and can be evoked through sharing of personal experiences in stories. While empathy is influenced by narrative content, intuitively, people respond to the way a story is told as well, through narrative style. Yet the relationship between empathy and narrative style is not fully understood. In this work, we empirically examine and quantify this relationship between style and empathy using LLMs and large-scale crowdsourcing studies. We introduce a novel, theory-based taxonomy, HEART (Human Empathy and Narrative Taxonomy) that delineates elements of narrative style that can lead to empathy with the narrator of a story. We establish the performance of LLMs in extracting narrative elements from HEART, showing that prompting with our taxonomy leads to reasonable, human-level annotations beyond what prior lexicon-based methods can do. To show empirical use of our taxonomy, we collect a dataset of empathy judgments of stories via a large-scale crowdsourcing study with N=2,624 participants. We show that narrative elements extracted via LLMs, in particular, vividness of emotions and plot volume, can elucidate the pathways by which narrative style cultivates empathy towards personal stories. Our work suggests that such models can be used for narrative analyses that lead to human-centered social and behavioral insights.

Development of interesting and complex characters is the most important element of a narrative. Presented in this work is fAIble II, an automated narrative generation system that focuses on character development. fAIble II leverages a graph database, containerized modules, knowledge templates, and language structuring to produce diverse and coherent stories. Story progression is driven by character perception, emotion, personality, and interaction with the story world. The resultant system has been tested via anonymous questionnaire. Responses suggest its ability to create diverse, sensible narratives using character development.

With Google, Microsoft and Facebook all pushing artificial intelligence, AI is becoming the next battleground for agencies, perpetually on the hunt for new service lines. AI basically gives machines the ability to think like humans. A simple example: You can have a one-on-one conversation with another person, but AI can talk to 500 people at the same time and make decisions based on real-time data to learn what's going on in each conversation, explained Dave Meeker, vp of Isobar's U.S. operations. In the context of advertising and marketing, AI theoretically means more personalized and interactive consumer experience, including targeted programmatic ad buys, identification of site visitors' decision-making patterns, conversational commerce like bots, as well as smarter search and recommendation engines on websites, according to six agency executives interviewed for this article. At the moment, with the help of AI developed by big tech companies, agencies are able to serve cognitive ads and integrate voice-activated assistants in their campaigns.

With Google, Microsoft and Facebook all pushing artificial intelligence, AI is becoming the next battleground for agencies, perpetually on the hunt for new service lines. AI basically gives machines the ability to think like humans. A simple example: You can have a one-on-one conversation with another person, but AI can talk to 500 people at the same time and make decisions based on real-time data to learn what's going on in each conversation, explained Dave Meeker, vp of Isobar's U.S. operations. In the context of advertising and marketing, AI theoretically means more personalized and interactive consumer experience, including targeted programmatic ad buys, identification of site visitors' decision-making patterns, conversational commerce like bots, as well as smarter search and recommendation engines on websites, according to six agency executives interviewed for this article. At the moment, with the help of AI developed by big tech companies, agencies are able to serve cognitive ads and integrate voice-activated assistants in their campaigns.

At Airbnb, we are focused on creating a place where people can belong anywhere. Part of that sense of belonging comes from trust amongst our users and knowing that their safety is our utmost concern. While the vast majority of our community is made up of friendly and trustworthy hosts and guests, there exists a tiny group of users who try to take advantage of our site. These are very rare occurrences, but nevertheless, this is where the Trust and Safety team comes in. The Trust and Safety team deals with any type of fraud that might happen on our platform. It is our main objective to try to protect our users and the company from various types of risks.

With robotic research on the rise, the implementation of artificial intelligence is paving the way for robotic achievements. An AI passed the first round of a literary competition, begging the question, how safe are the creative arts? After an artificial intelligence software proved creative enough to -- with the help of humans -- pass the first round of a national Japanese literary competition, people have now become more afraid of AI. Of course, now AI has seeped into the arts, so there's no stopping them. We're going to be taken over by computers, right?