The penetration of artificial intelligence (AI) into our daily lives over the last several years has been rather incredible. Siri, Alexa, Cortana, Waze, Google assistant, music recommendations – the list goes on. The ability to talk to devices – and have them respond – used to be science fiction not too long ago. (Think HAL 9000, Skynet, Ultron… wow, Hollywood really likes its AIs to be evil!) Now it is common to interact with an AI device on a daily basis.

AI in Higher Education

Many institutions of higher education have started utilizing AI devices as student support tools in dorm rooms and around campuses. (Read about the Alexa project at St. Louis University.) We see the next critical step in the evolution of the use and value of these devices is getting them in the hands of university administrators.

Over the last decade, the amount of data available at institutions has grown immensely. Though this could be a “chicken or the egg” argument, making informed decisions based on that data has also grown. Not only is more data available today than in the past, but the data is more readily accessible. (And it is usually accessible with some type of reporting tool or visualization aid.) More departments are using more data to make better decisions.

Tech Companies and Higher Ed

The major firms (Amazon, Google, Oracle, Microsoft to name a few) all have various initiatives targeting various aspects of higher education. From recruiting to giving, all functional areas are receiving some level of attention. These initiatives are good and there are some impressive results.

The issue with many of these tools is that you need to be an expert at data visualization, a data scientist, or a database expert to effectively use any of them. Even the “simple” ones require at least some level of technical skill and/or knowledge.

The Issue at Hand

The issue of getting data into the hands of the appropriate end users in a timely manner has been improved but not solved. The end user still needs to wait for some expert to produce the data or navigate some arcane tool to get the answer they are seeking. Often, the data cannot be accessed in an easy manner because it is hidden behind some fog bank of technical obscurity. Or, it’s only available if the user is fluent in ancient Greek. (I kid. Sort of.) Unfortunately for most end users, these situations are all too real, and thus the promise of ubiquitous and seamless data access is not achievable under current circumstances.

Say “Hi” to Voice AI

Enter the voice user interface. If the end user can ask for information and a system can be crafted to understand what the user is asking for – progress can be made. There are certainly challenges that need to be overcome. Still, initial attempts show promise in the ability to provide fast, accurate answers to important questions. And the questions are framed by the end user, not in some weird language known only to a few.

The querying of data from data sets, and the presentation of that data, using a voice interface is just in the beginning stages. Soon to follow this data discovery phase will be data analysis, trend identification, and even data visualization. Vendors are already working on predictive analysis and recommendation engines.

Expectations

Voice interfaces must respond to input reliably or they will be rejected by users – and eventually by higher education as a whole. Quite a bit of work still needs to be done in this realm. But that was also true of Siri and her cohorts when they first appeared. Now look at them today.

The performance of the application must be fast as well. Complex data analysis takes time, even for the fastest processors. End users will not want to wait long for an answer after asking a question. Even with advanced development tools, constructing an effective voice interface requires an in-depth understanding of both the tasks to be performed and the target audience that will use the final system. The closer the voice interface matches the user’s mental model of the task, the easier it will be to use with little or no training. This will result in both higher efficiency and higher user satisfaction.

This research and development in higher education is proceeding quickly, but by no means is it finished. Look for voice to become the dominant method of asking questions and interpreting data in the coming years. When voice output is eventually paired with visual displays, the holy grail of information presentation may finally be achieved.

Conclusion

So, what is the key to success when it comes to voice AI in higher ed? It is to ensure that the end user can arrive at the same conclusion using the same data and same methods of calculation that they would if using a more “standard” reporting tool.

Transparency of both the data and the process will build the trust necessary for these systems to succeed. That is no different than how things are today. Even with the best data reporting and visualization tools today, the data and the process for the calculations must be discoverable by the end user. If the results are not trusted by the end user, the software quickly becomes “shelf-ware.”