How Krashen's Hypotheses Are Still Pivotal In iCALL?
For decades, one of the foundational theories driving the area of linguistics known as Second Language Acquisition (SLA) has been that of Stephen Krashes’s Input Hypothesis (which is really 5 interconnecting hypotheses), and the accompanying formula which states that, given the learners level of language, say i, the learner will improve if and only they are exposed to language just one step beyond their current level of competence, hence the formula i + 1.
While theoretically elegant, the practical application is a nightmare, especially in the classroom... how does a single teacher provide tailored "comprehensible input" to a massive cohort of students!?
This is where modern Intelligent Computer-Assisted Language Learning (dubbed iCALL) is attempting to solve this through the rise of voice-first conversational AI.
The Push For Comprehensible Output
Building on Krashen focus on input, researchers like Merrill suggest that effective instruction requires Application. This is where apps like Pingo AI and the voice conversation features in Duolingo change the game. By forcing learners into real-time voice interactions, these apps move beyond passive listening into pushed output.
As noted in systematic reviews of iCALL instructional design, the key to success is providing immediate, personalized feedback post lesson. When you speak to a bot on Pingo AI, the system isn't just listening; it’s calculating your "i" level in real-time to adjust its responses, storing it, and then the following session building on it at the learner’s pace.
The Semantic Gap: Understanding vs. Transcribing
However, there is a massive technical hurdle in how these apps "hear" us. A way in which Automatic Speech Recognition (ASR) systems are evaluated is by, focusing on Word Error Rate (WER), meaning how many words are wrong in between what has been said, and what the system has understood. For English, if the system misunderstands one in every twenty words, it is considered top quality. This can cause a losing of trust between the learner and technology, especially in scenarios where the learner truly knows that what they’re saying is correct, but are being marked wrong.
Additionally, the system checks to see if you’ve matched the exact words in the exact order according to what the it believes the right answer to be, which for a language learner can be discouraging, and even more so when the phrase is considered correct to a native speaker of the language, but simply, the learner chose to use a different set of synonymous words.
Because of this, recent research suggests that we should be moving toward a new metric called Semantic Distance, where if a learner uses a dialectal variant of a word/phrase that still carries the correct meaning, the system should be smart enough to recognize the success of the communication, even if it doesn’t match the answer that the system has saved internally. This is vital for maintaining the flow of learning, as by not discouraging the learner to learn other dialects of the language.
Lowering The Affective Filter
Perhaps the biggest advantage of apps like Duolingo is their impact on the Affective Filter. Krashen argued that high anxiety acts as a mental block to acquisition. Most learners are terrified of speaking to native speakers for fear of judgment.
Apps provide a "safe" space where the "bot-as-tutor" model can lower this filter significantly. However, there is a “gamification trap.” While levels and streaks help beginners, advanced datasets and balanced corpora show that for more proficient speakers, the "game" can feel restrictive. To truly reach an advanced i + 1 level, the conversation must feel authentic and problem-centred, rather than a series of gamified drills.
In Summary
The future of CALL isn't just "smarter" apps, it’s apps that understand the human element of learning. By focusing on semantic meaning and providing a low-anxiety environment for "pushed output," platforms like Pingo AI and Duolingo are finally making the i + 1 formula a scalable reality.