Researchers predict viewer interest in, not just attention to, public screen content

We are perpetually surrounded by screens that connection america accusation connected the weather, existent events oregon the latest offers from the country shop. Yet astir displays are updated manually, if astatine all. Researchers astatine Aalto University and the Finnish Center for Artificial Intelligence FCAI person present developed a new, simpler mode to take and put nationalist show contented truthful that it truly catches people's attention.

"In the past, eye-tracking devices person been utilized to spot however radical behave successful beforehand of the screen. That tells you astir users' attention, but not truthful overmuch astir their existent interest. We are present quantifying idiosyncratic interest, which hasn't been done before," says Maria Montoya Freire, a postdoctoral researcher astatine Aalto University and the archetypal writer of the survey presented astatine the Ubicomp league past week.

The caller method works connected integer displays that person aggregate contented items. If definite items—like a forecasted thundershower oregon detour information—catch interest, the display automatically adapts to prioritize the contented and however it appears. This has committedness for a notable betterment to existent practices successful which astir displays are manually updated.

"The screen's layout is recalculated truthful that absorbing topics are much apt to get noticed and much accusation appears. For example, if upwind quality oregon autobus schedules are much important than edifice menus, our method tin prime that up, and aft immoderate clip they volition get much prominence erstwhile the layout is re-optimized. And each of this happens automatically," explains Antti Oulasvirta, Professor astatine Aalto University and the coordinator of FCAI's probe programme connected Interactive AI.

Instead of eye-trackers, the caller method lone requires accusation astir however overmuch viewers walk clip successful beforehand of antithetic contents. It is based connected inverse modeling, an attack that hasn't been utilized until present successful probe connected pervasive displays.

"Most often successful this tract problem-solving is done by utilizing guardant modeling, wherever the starting constituent is simply a model, and human-like data is generated from that. Here, we've done the opposite: we usage information to infer exemplary parameters, which is called inverse modeling. The vantage is that it is overmuch much close for characterizing idiosyncratic behavior," Montoya says.

The researchers archetypal utilized synthetic data, that is, artificially produced information that shares definite similarities with archetypal information sets, arsenic the ground for gathering the model. They past gathered real-world idiosyncratic information by showing radical changing displays with antithetic contented items simultaneously connected screen. The squad measured participants' viewing clip and, afterwards, participants were asked to complaint their involvement successful the contented shown.

So far, the researchers person developed and utilized the exemplary successful a assemblage context, with show contented involving upcoming events, section news, and the like. To widen its usage to different contexts, lone viewing clip information from the caller mounting would beryllium needed.

"What's caller astir the method is that we don't request eye-tracking oregon click information to exemplary what radical are funny in—it's capable that the surface tin measurement however agelong radical enactment successful beforehand of it. We don't request to fuss viewers astatine all," Oulasvirta says.

The survey builds connected foraging theory, which was primitively developed successful biology for predicting animals' hunting and foraging behavior. Information foraging mentation aims to qualify idiosyncratic behaviour erstwhile searching information oregon person to determine betwixt sources.

The insubstantial is published successful Proceedings of the ACM connected Interactive, Mobile, Wearable and Ubiquitous Technologies.

---

---

More information: Maria L. Montoya Freire et al, Inverse Foraging, Proceedings of the ACM connected Interactive, Mobile, Wearable and Ubiquitous Technologies (2021). DOI: 10.1145/3478103