Loora

Put People First

• Loora is designed to tackle an important use case for English language learners (ELLs), as it aims to help learners practice the kind of conversational English used in everyday life. This can give people a private and nonjudgmental space to focus on conversations about the topics that matter most to them.
• Achieving fluency in any language happens best in an immersive environment. But, while many ELLs are in those environments at work or school, they don't always have the benefit of feedback and instruction during their day-to-day conversations. This has left ELLs with few choices, often having to rely on expensive, one-on-one tutors with limited availability. If products like Loora are able to fill this gap, it could be highly beneficial for ELLs around the world.
• Loora promotes a positive learning environment by praising learners for their English skills.
• Loora lets speakers continue without interruptions, and only addresses mistakes and corrections once the speaker is finished.
• Loora scores pronunciation in two ways: word by word, and by each full response. Loora seems to make several assumptions about what constitutes correct pronunciation. Even as fluent English speakers, our reviewers received surprisingly low pronunciation scores. It's essential for users to understand that Loora's scoring represents just one pronunciation option among many.
• During the signup process, learners can specify their English proficiency level: beginner, basic, intermediate, advanced, or professional. If a learner chooses "beginner," the subsequent prompts and signup screens will still be presented in English. Genuine beginners might struggle with the system's English instructions.

Be Effective

• Loora does not interrupt the flow of conversation with its chatbot. Instead, it highlights where it assesses grammar and pronunciation errors as it transcribes speech. These corrections and scores are then available for learners to dive into at their own convenience. Loora's specific, actionable, and timely feedback offers both correction and encouragement, and is an educational best practice.
• Loora offers a translation feature that allows learners to type their desired phrases and receive an audio + text translation in English. This enables them to practice words and expressions they may not yet know. We do not have information on the extent of Loora's non-English language support.
• We do not know how well Loora serves English language learners with speech impairments or disabilities.

Prioritize Fairness

• While we cannot say for certain whether Loora is specifically targeting the kinds of unfair treatment that ELLs can experience, it has a unique potential to do so. For example, language-based stigma targeting ELLs has bee shown more negative interactions with others. In other cases, non-fluent speakers using the same scripted responses as fluent speakers were less likely to be recommended for middle-management positions or to receive venture funding. Just being classified as an ELL student has been shown to have a direct negative impact on test scores, high school graduation rates, and college attendance.
• Currently, Loora is more affordable than a private English tutor and is available 24/7.
• While not consistent, Loora's responses were often inaccurate and at times reinforced unfair biases in our testing.
• Loora's user interface seems to put a large emphasis on pronunciation scoring and correction. Fluency in grammatically correct English doesn't require a uniform accent for every individual. When it comes to being understandable, grammar carries more weight in real life than pronunciation.
• While generally more affordable than a human tutor, $120/year or $15/month could still be out of reach for many ELLs.
• Loora is built on multiple large language models (LLMs). While we don't have information about the training data used to power Loora, it includes text that is publicly available on the internet. This data is more likely to represent the internet-connected population, which in turn means it overrepresents people in wealthier nations, as well as views from people who are wealthier, younger, and male. In other words, the quantity of training data does not guarantee its diversity. Loora strives to customize conversations according to learners' selected interests, encompassing a wide range of subjects, from art and culture to science. Yet, given the diverse backgrounds of ELLs, there's uncertainty regarding Loora's ability to effectively cater to such a broad spectrum of culturally specific interests.

Help People Connect

• Loora can help to foster more self-assuredness in speaking English, in turn enabling more meaningful human contact and connection.
• Loora is available 24/7 and provides a safe, pressure-free space to practice English.
• Pronunciation scores, particularly if consistently low, might undermine one's confidence in their ability to communicate effectively with real people.

Be Trustworthy

• Loora is intended for users age 16+, which protects younger users from the risks inherent in large language models (LLMs). If a younger learner, under 16, attempts to sign up, their emails are added to a blocklist.
• Loora does not provide information on how it determines its pronunciation scores. Without this, users cannot know whether this scoring is grounded in learning science.
• Loora is built on multiple LLMs. All LLMs are each challenged by biased outputs, inaccuracies, and "hallucinations"—an informal term used to describe the false content or claims that are often output by generative AI tools. At the time of writing, we are not aware of if or how Loora has fine-tuned or safeguarded its use of these models.

Use Data Responsibly

• It does not appear possible for learners to control and delete their data.
• Other chatbots built on LLMs specify that users are responsible for not putting personal or confidential information into the chatbot conversations to prevent the chatbot from training on personal information. Loora does not provide any such warnings.
• Loora's Terms of Service indicate that content provided by users—including text and audio recordings—can be used and shared by Loora in many ways, including to "create the Generated Content." Neither these terms nor the company's privacy policy specifically indicate whether learner data is used to create or train their models. The company's terms give Loora the right to use and "make available to the public" photos, texts, and audio recordings created while using Loora to create—and then own—"generated content." Notably, public interviews with Loora's creators indicate that they do train their models "on their own data."

This review is distinct from Common Sense's privacy evaluations and ratings, which evaluate privacy policies to help parents and educators make sense of the complex policies and terms related to popular tools used in homes and classrooms across the country.

Keep Kids & Teens Safe

• While Loora's sign-up process has an age gate, there is nothing to stop kids from signing up if they choose to give an incorrect birth date.

Be Transparent & Accountable

• Loora does not provide warnings that the chatbot may produce biased, incorrect, harmful, or misleading content.
• Loora does not provide a means of giving feedback to the system if a user believes the outputs are inaccurate.
• At the time of this review, Loora has not shared information about what powers its system. Public information indicates that Loora initially built its own LLMs, but is now leveraging a "variety" of available LLMs for its service, while continuing to train on the company's own data. Without knowing which LLMs it is using or having information about how it is either fine-tuning those or creating its own data sets, users are in the dark about whether Loora is evaluating its system for accuracy or fairness. Users also have no information about how the company might be trying to limit the chatbot from generating harmful content.
• Because chatbots based on LLMs can generate responses that sound correct and authoritative but are not, it can be easy for users to take this information at face value.