Human detection of political speech deepfakes across transcripts, audio, and video

Nature Communications, Год журнала: 2024, Номер 15(1)

Опубликована: Сен. 2, 2024

Abstract Recent advances in technology for hyper-realistic visual and audio effects provoke the concern that deepfake videos of political speeches will soon be indistinguishable from authentic video. We conduct 5 pre-registered randomized experiments with N = 2215 participants to evaluate how accurately humans distinguish real fabrications across base rates misinformation, sources, question framings without priming, media modalities. do not find misinformation have statistically significant on discernment. deepfakes produced by state-of-the-art text-to-speech algorithms are harder discern than same voice actor audio. Moreover all framings, we information enables more accurate discernment text alone: human relies something is said, audio-visual cues, what speech content.

Язык: Английский

---

Chatting with a Learning Analytics Dashboard: The Role of Generative AI Literacy on Learner Interaction with Conventional and Scaffolding Chatbots

Опубликована: Фев. 21, 2025

Язык: Английский

---

To Rely or Not to Rely? Evaluating Interventions for Appropriate Reliance on Large Language Models

Опубликована: Апрель 24, 2025

Язык: Английский

---

Building Proficiency in GAI: Key Competencies for Success

Опубликована: Янв. 17, 2025

The rapid proliferation and adoption of generative Artificial Intelligence (GAI) underscores its ease use. However, there has been limited research exploring what constitutes proficient use GAI competencies underpin it. In this study, we adopt a grounded approach semi-structured interviews to explore how twenty-five expert users (all knowledge workers) define, exemplify, explain proficiency. A purposive sampling was adopted with the aim capturing input from experts range occupations sectors towards answering three questions. First, can identify characteristics that differentiate (more effective) GAI? Second, are seen underlie Third, benefits associated more tools? Analysis descriptions shared by revealed four aspects proficiency: effective prompting, informed responsible choices, diversity use, complexity frequency addition, following themes emerged analysis supporting GAI: literacy, domain expertise, communication skills, metacognition curiosity inquisitiveness, flexibility adaptability, diligence, (in some contexts) information technology skills. More have ranging improved productivity, higher quality output, original work. By offering comprehensive framework for GAI, in real-world experience, study guides further substantiates continuing relevance human knowledge, mindsets when working tools.

Язык: Английский

---

An Automated Hierarchy Method to Improve History Record Accessibility in Text-to-Image Generative AI

Applied Sciences, Год журнала: 2025, Номер 15(3), С. 1119 - 1119

Опубликована: Янв. 23, 2025

This study aims to enhance access historical records by improving the efficiency of record retrieval in generative AI, which is increasingly utilized across various fields for generating visual content and gaining inspiration due its ease use. Currently, most AIs, such as Dall-E Midjourney, employ conversational user interfaces (CUIs) creation retrieval. While CUIs facilitate natural interactions between complex AI models users making process straightforward, they have limitations when it comes navigating past records. Specifically, require numerous interactions, must sift through unnecessary information find desired records, a challenge that intensifies volume grows. To address these limitations, we propose an automatic hierarchy method. method, considering modality characteristics text-to-image applications, implemented with two approaches: vision-based (output images) prompt-based (input text) approaches. validate effectiveness method assess impact approaches on users, conducted 12 participants. The results indicated enables more efficient than traditional CUIs, preferences varied depending their work patterns. contributes overcoming linear existing CUI systems development It also enhances accessibility, essential function effective tool, suggests future directions research this area.

Язык: Английский

---

Prompting Theory into Practice: Utilizing ChatGPT-4 in a Curriculum Planning Course

Education Sciences, Год журнала: 2025, Номер 15(2), С. 196 - 196

Опубликована: Фев. 6, 2025

As of late, generative AI tools have been rapidly gaining purchase as an important part life. Thus, one cannot ignore their growing integration into educational landscapes, including teacher education. This qualitative study examines how pre-service teachers (PSTs) leverage ChatGPT-4 to apply constructivist theory in curriculum planning (CP). The findings revealed three approaches through which PSTs used the chatbot practice: (a) simplifying theory, (b) applying and (c) visualizing theory. suggest that need refine prompts using curricular language engaging creative critical thinking supported translation process. In incorporating CP, considered multiple factors, ideation, inspiration, creativity, reliability, insufficient personalization—attesting a balanced perspective on use this tool, i.e., recognizing potential benefits utilizing chatbot, while remaining cognizant its associated risks limitations. points aspects CP AI, educators should discuss with PSTs.

Язык: Английский

---

Fostering Appropriate Reliance on Large Language Models: The Role of Explanations, Sources, and Inconsistencies

Опубликована: Апрель 24, 2025

Large language models (LLMs) can produce erroneous responses that sound fluent and convincing, raising the risk users will rely on these as if they were correct. Mitigating such overreliance is a key challenge. Through think-aloud study in which participants use an LLM-infused application to answer objective questions, we identify several features of LLM shape users' reliance: explanations (supporting details for answers), inconsistencies explanations, sources. large-scale, pre-registered, controlled experiment (N=308), isolate effects reliance, accuracy, other measures. We find presence increases reliance both correct incorrect responses. However, observe less when sources are provided or exhibit inconsistencies. discuss implications findings fostering appropriate LLMs.

Язык: Английский

---

Facilitating Generative AI Literacy in the Face of Evolving Technology: Interventions in Marketing Classrooms

Journal of Marketing Education, Год журнала: 2025, Номер unknown

Опубликована: Фев. 18, 2025

The emergence of generative AI (GenAI) has illustrated that higher education needs to adapt the technology. Its speed evolution requires we adequately prepare students for an ever-changing landscape. Toward achieving aim, draw on concept interpretive flexibility, where interpretations, uses, and outcomes a new technology can differ evolve over time, often with dominant stakeholders controlling process. To engage marketing in this process, propose they be presented these diverse interpretations now as part GenAI literacy. Specifically, offer three small-scale pedagogical interventions designed address urgent need. Given newness GenAI, our are infused into existing instruction, instead requiring redesign curriculum. With each intervention, not only significantly decrease their confidence accuracy what produces but also see reasons examine implications it. Both outcomes, suggest, could help maintain flexibility required properly respond guide its impacts, become evident. We encourage educators prioritize comprehensive notion literacy pedagogy flexibility.

Язык: Английский

---

Building Proficiency in Generative AI: Key Competencies for Success

Опубликована: Фев. 20, 2025

The rapid adoption of generative Artificial Intelligence (GenAI) underscores its ease use, yet research on GenAI proficiency and competencies is limited. This study uses semi-structured interviews with twenty-five expert users from various sectors to explore proficiency. aims answer three questions: What differentiates proficient use? support benefits does use provide? Three aspects emerged: effective prompting, informed responsible choices, diverse, complex use. following were seen GenAI: literacy, domain expertise, communication skills, metacognition, curiosity, flexibility, adaptability, diligence, IT skills. outcomes improved productivity, higher quality output, greater originality. framework, grounded in real-world experience, the importance human knowledge, mindsets for tools.

Язык: Английский

---

Unlocking Proficiency: Experts’ Views on the Use of Generative AI

Опубликована: Март 25, 2025

The rapid proliferation and adoption of generative Artificial Intelligence (GenAI) underscore its ease use. However, there has been limited research exploring what constitutes proficient use GenAI competencies underpin it. In this study, we used semi-structured interviews to explore how twenty-five expert users (all knowledge workers) define, exemplify explain proficiency. A purposive sampling approach was adopted with the aim capturing input from experts a range occupations sectors towards answering three questions. First, can identify characteristics that differentiate (more effective) GenAI? Second, are seen underlie Third, benefits associated more tools? Analysis descriptions shared by revealed four aspects proficiency: effective prompting, informed responsible choices, diversity complexity use, frequency addition, following themes emerged analysis supporting GenAI: literacy, domain expertise, communication skills, metacognition curiosity inquisitiveness, flexibility adaptability, diligence (in some contexts) information technology skills. More have ranging improved productivity, higher quality output original work. By offering comprehensive framework for GenAI, grounded in real world experience, study guides further substantiates continuing relevance human mindsets when working tools.

Язык: Английский

---