Thanks for watching Lawrence. Please let us know if you have any questions. It was a wonderful collaborative experience.

Thanks for watching Eric. Elva was inspired, in part, by the NSF support of the Finland research exchange you directed.

These are great questions. We trained the system so that there was always a default response (e.g., the definition of the word). There was a training period that required us to make guesses of what students would say and based on that decide what would be the most appropriate response. The more data we collected, the better the system became in recognizing student speech. There are limitations to the system we used in terms of language understanding that we tried to offset by having the avatar use more generic prompts. Despite the errors in speech recognition, most students were willing to "forgive" the avatar when they felt she was not understanding them. I hope this helps!

Great questions Andee. I was the lead developer of the science content and the funding for my doctoral studies came from this grant. To follow-up on Doris' comments, an innovative part of the system was the accuracy of the word recognition, that allowed us to use the platform. The system is not true artificial intelligence, because the responses were not necessarily natural. We developed the content based on many hours of data collected from second-grade speech patterns, possible responses the students would give, and various grade-level appropriate discourse. The purpose was to guide their thinking towards a specific vocabulary development using different prompts. To account for the correct answers, we determined various iterations that students could use and decided what was an "acceptable" or "correct" answer based on the research we had about the vocabulary. We tried to develop a more appropriate form of AI, but quickly found that we were limited by the technology and what's available in terms of more adequate response recognition that would allow us to make a system that is easy to modify. Hope this helps Andee. Let me know if you have further questions.

Thank you so much Erica for your positive comments. The main difference with ELVA and other language learning programs is that it targets vocabulary that students will see during whole group instruction. So the content is aligned with second grade science and social studies lessons. The goal is that students practice the vocabulary that they can then use during whole group instruction or when talking in small groups about science and social studies. We have had several teachers tell us that students who participated in ELVA got very excited when they heard the words taught during whole group instruction. Not only that, they shared the definitions they had learned with the rest of the class, they used the words learned when answering questions, and they appeared to be more engaged in the whole group lessons.

Thank you Kelsey. Yes, watching students making the connections of the words to other contexts was very exciting. Students have not told us that they feel safer speaking in English with the avatar, but as you saw on the video, they were much more willing to take risks and express their ideas even if they might not have been completely fleshed out. Some students responded in Spanish too which lead us to believe that they felt confident in responding the best way they could.

That's wonderful, thank you Doris!

Great questions Mari. We used an adapted version of a system developed by Boulder Learning. We believe that the words would be relevant across the country because we focused on words that would address the science content suggested by the Next Generation Science Standards, CCSS, and the TEAKS. Most of the words were abstract Tier 2 words (see Beck & McKeown, Snow) that would be relevant to discuss academic content and that we know textbooks don't always explain in-depth (e.g., impact, balance, erupt). Yes, we are looking into a system now that will allow us to create lessons faster, and also that will allow teachers eventually to create their own vocabulary lessons.

Hi Mari, I was the lead content developer for science and worked on the project since the beginning. To follow-up on Doris' comments, the program was not true AI where the responses were totally adaptable to "every" possible circumstance. Rather we created possible prompts and answers that were collected through interactions with the avatar. The initial idea was to create a form of AI but we were limited by the technology available to us. However, as we continue to advance in the development of more natural forms of AI á la Siri or Alexa and as they become more accessible to us, we can see that this method of teaching vocabulary in a deliberate and structured form is a promising way to teach in-depth. Hope this helps Mari and let me know if you have further questions.

Hi Garima. Thanks for sharing your interesting approach to teaching science with us. The idea of using cards with both languages that are available to teachers and students is very good and it provides an opportunity for both groups to improve their bilingualism. In our study, we had students who responded in Spanish to the prompts because they were learning Spanish in a dual language program most of the day. Our findings indicated that students learned the academic words the avatar taught them independently of their English language proficiency. Now we are in the process of analyzing the quality of student responses to see if their responses in Spanish differed from the responses in English. So far, we have not found a difference.

Hello Lawrence and thanks for your comments. I was the lead developer of the science lessons and the grant from ELVA helped fund my doctoral studies with Dr. Baker. One of the most revealing things that we have learned with the program development and the pilot studies is that this program can serve as a platform, not only for other languages but also for other subjects as it helps in the development of vocabulary in-depth. In the future, we hope to extend it beyond 2nd grade as a continuation of the learning process.

Hello Paul, I've recommended your video to one of our graduate students who has developed a program for teaching Inupiaq vocabulary to students on Alaska's North Slope.  That program has been positive, but there's always the challenge of getting these students, who are learning Inupiaq as a second language, to construct sentences and use the vocabulary in speech.  That part would be very different from working with Spanish speaker in the USA who have much exposure to English, since Inupiaq is present in the students' lives but not to an extent that causes them to learn to speak it, in most cases.  This is the plight of many small languages in this country and even worldwide. 

Thanks so much for recommending our video, Lawrence. I agree with your comment referring to the exposure perspective. I firmly believe in the value of preserving a language because by doing so we preserve more than just language. I have seen some similar work on small language preservation in Mexico where it is institutionalized through textbooks and teaching strategies for indigenous groups, yet this is at a small scale. Keep up this much needed work!

Hi Alison. Thanks for watching our video. Regarding your question about the measures, yes, we used also a measure where students had to use the target word in a sentence. In addition, we had several standardized measures. The results on our standardized measures were not significant, but the effect sizes were moderate. We also introduced every target word in an appropriate context, and we created comprehension questions where students had to respond using the words taught. Finally, the program includes activities  (multiple choice, yes/no, building sentences) were students have to explain their answers. We are in the process of transcribing and analyzing their responses.

Teachers indicated that students participated much more in their lessons after they had been exposed to the ELVA words. Some of them provided the vocabulary definitions and also explanations of the science topic teachers were covering. We had anticipated this because words were selected taking the Next Generation Science standards, CCSS, and TEKS into account. We are exploring now, however, better ways to connect the vocabulary lessons to teachers regular instruction so they can use the information from ELVA to also guide their instruction.

Chih-Che,

Thank you for watching our video. Yes, we plan to have an app-like version that is compatible with Windows and Mac. In fact, we have already begun to merge the content into a web-based classroom management software system to pilot it. Our intention is that this will make ELVA accessible to more students and allow participation both in and outside of school. We look forward to seeing how students respond.

Thank you so much for watching Christina. Your comment reflects exactly what we attempted to do on this project. Students could practice using their language skills without worrying about making mistakes. We found that at first, some of the shy students would not say very much, but once they realized that it was ok to speak, and that they actually had the answers to many of the questions the avatar asked them, then they started talking more and looking forward to interacting with ELVA. We will continue researching ways to maximize student opportunities to develop their language proficiency through technology. We will keep you posted!!